[libnbd PATCH 00/23] continue wrapping generated C code harder
Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=2172516 Repo: https://gitlab.com/lersek/libnbd.git Branch: wrap-2172516-wave4 This series is a continuation of the one noted in <https://bugzilla.redhat.com/show_bug.cgi?id=2172516#c15>. It reduces the widths of the following generated C files: file width before width after ------------------ ------------ ----------- include/libnbd.h 81 79 lib/api.c 206 79 ocaml/nbd-c.c 108 79 python/libnbdmod.c 130 79 python/methods.h 104 80 The following generated C files remain for the next wave: file width ------------------ ----- lib/states-run.c 102 lib/states.c 116 lib/states.h 123 Quite a few patches in the series are easier to review with "git show -b" or "git show --color-words", due to code reindentation or other relatively mechanical changes. (And I've just found out about "git show --color-moved", which works really well on patch #08, 'generator/utils: factor out "pr_buf"', for example.) Most patches where such display tweaks are beneficial mention that fact. These formatting tweaks are also the reason why I've pushed the branch to my libnbd fork on gitlab (see URL & branch name above). The posted patches should be possible to apply with git-am, for merging (thus, they couldn't be formatted with the "-b" option, for example), but for review, it may be easier to just fetch them (or to fetch them in addition) and to show them locally. Thanks for reviewing, Laszlo Laszlo Ersek (23): generator: split LIBNBD_ATTRIBUTE_NONNULL replacement text generator: wrap "python/methods.h" at 80 characters generator: wrap "python/libnbdmod.c" at 80 characters generator: split "int64_from_uint32_array" calls in "ocaml/nbd-c.c" generator: wrap nbd_internal_ocaml_raise_closed() calls in "ocaml/nbd-c.c" generator: factor out "num_args" in "print_ocaml_binding" generator: wrap byte-code compat function bodies in "ocaml/nbd-c.c" generator/utils: factor out "pr_buf" generator/utils: comment on "maxcol" handling in "pr_wrap" generator/utils: add "pr_wrap_cstr" generator/C: print_trace_enter: break "enter" to a common new line generator/C: print_trace_enter: expand "function" shorthand in value iters generator/C: print_trace_enter: print Closure values uniformly with others generator/C: print_trace_enter: generate commas as suffixes, not prefixes generator/C: print_trace_enter: strip spaces around PRI* macros generator/C: print_trace_enter: wrap "enter" debug() / debug_direct() call generator/C: print_wrapper: break apart the "ready" & "processing" states generator/C: print_wrapper: fold permitted states error message generator/C: print_trace_leave: separate arg list from function designator generator/C: print_trace_leave: wrap "leave" debug() / debug_direct() call generator/C: print_flags_check: break guard condition to a new line generator/C: print_wrapper: use helper variable for permitted state check generator/C: lib/api.c: indent arg list 2 spaces relative to function name generator/C.ml | 195 ++++++++++++-------- generator/OCaml.ml | 27 ++- generator/Python.ml | 13 +- generator/utils.ml | 85 ++++++++- generator/utils.mli | 1 + ocaml/helpers.c | 2 +- ocaml/nbd-c.h | 3 +- 7 files changed, 225 insertions(+), 101 deletions(-) base-commit: 0744f748ed900fb0537da9a5b6532538f3c78e83
The LIBNBD_ATTRIBUTE_NONNULL replacement text is just a bit too long; split the macro definition similarly to LIBNBD_ATTRIBUTE_ALLOC_DEALLOC. It changes "include/libnbd.h" like this:
#ifndef LIBNBD_ATTRIBUTE_NONNULL #if defined (__GNUC__) && LIBNBD_GCC_VERSION >= 120000 /* gcc >= 12.0 */ -#define LIBNBD_ATTRIBUTE_NONNULL(...) __attribute__ ((__nonnull__ (__VA_ARGS__))) +#define LIBNBD_ATTRIBUTE_NONNULL(...) \ + __attribute__ ((__nonnull__ (__VA_ARGS__))) #else #define LIBNBD_ATTRIBUTE_NONNULL(...) #endif
Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=2172516 Signed-off-by: Laszlo Ersek <lersek@redhat.com> --- generator/C.ml | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/generator/C.ml b/generator/C.ml index 61ca43d95db5..b65d3de2e82d 100644 --- a/generator/C.ml +++ b/generator/C.ml @@ -420,7 +420,8 @@ let pr "\n"; pr "#ifndef LIBNBD_ATTRIBUTE_NONNULL\n"; pr "#if defined (__GNUC__) && LIBNBD_GCC_VERSION >= 120000 /* gcc >= 12.0 */\n"; - pr "#define LIBNBD_ATTRIBUTE_NONNULL(...) __attribute__ ((__nonnull__ (__VA_ARGS__)))\n"; + pr "#define LIBNBD_ATTRIBUTE_NONNULL(...) \\\n"; + pr " __attribute__ ((__nonnull__ (__VA_ARGS__)))\n"; pr "#else\n"; pr "#define LIBNBD_ATTRIBUTE_NONNULL(...)\n"; pr "#endif\n";
The argument list for those Python method declarations that can grow too wide is always "PyObject *self, PyObject *args"; simply break that fixed arg list to a new line. Effect demonstrated on the longest line in "python/methods.h":
-extern PyObject *nbd_internal_py_aio_connect_systemd_socket_activation (PyObject *self, PyObject *args); +extern PyObject *nbd_internal_py_aio_connect_systemd_socket_activation ( + PyObject *self, PyObject *args + );
Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=2172516 Signed-off-by: Laszlo Ersek <lersek@redhat.com> --- generator/Python.ml | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/generator/Python.ml b/generator/Python.ml index d493f17a5d6a..87444c54bdf0 100644 --- a/generator/Python.ml +++ b/generator/Python.ml @@ -70,8 +70,9 @@ let List.iter ( fun name -> - pr "extern PyObject *nbd_internal_py_%s (PyObject *self, PyObject *args);\n" - name; + pr "extern PyObject *nbd_internal_py_%s (\n" name; + pr " PyObject *self, PyObject *args\n"; + pr " );\n" ) ([ "create"; "close"; "display_version"; "alloc_aio_buffer";
Break the fields of each PyMethodDef structure initializer to separate lines. While at it, drop the superfluous (char*) cast from the "PyMethodDef.ml_name" field initializer <https://docs.python.org/3/c-api/structures.html#c.PyMethodDef>. (According to commit bf0eee111fc3, "Add Python 3 bindings.", 2019-05-11, we don't target Python 2.) Effect shown on the currently longest line in "python/libnbdmod.c":
- { (char *) "aio_connect_systemd_socket_activation", nbd_internal_py_aio_connect_systemd_socket_activation, METH_VARARGS, NULL }, + { + "aio_connect_systemd_socket_activation", + nbd_internal_py_aio_connect_systemd_socket_activation, + METH_VARARGS, + NULL + },
Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=2172516 Signed-off-by: Laszlo Ersek <lersek@redhat.com> --- generator/Python.ml | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) diff --git a/generator/Python.ml b/generator/Python.ml index 87444c54bdf0..c146f04f0b49 100644 --- a/generator/Python.ml +++ b/generator/Python.ml @@ -100,8 +100,12 @@ let pr "static PyMethodDef methods[] = {\n"; List.iter ( fun name -> - pr " { (char *) \"%s\", nbd_internal_py_%s, METH_VARARGS, NULL },\n" - name name; + pr " {\n"; + pr " \"%s\",\n" name; + pr " nbd_internal_py_%s,\n" name; + pr " METH_VARARGS,\n"; + pr " NULL\n"; + pr " },\n" ) ([ "create"; "close"; "display_version"; "alloc_aio_buffer";
Best explained with the diff that results in the generated "ocaml/nbd-c.c" source file:
--- old 2023-04-21 10:44:30.192602608 +0200 +++ new 2023-04-21 10:51:46.582658399 +0200 @@ -205,7 +205,10 @@ extent_wrapper_locked (void *user_data,
metacontextv = caml_copy_string (metacontext); offsetv = caml_copy_int64 (offset); - entriesv = nbd_internal_ocaml_alloc_int64_from_uint32_array (entries, nr_entries); + entriesv = nbd_internal_ocaml_alloc_i64_from_u32_array ( + entries, + nr_entries + ); errorv = caml_alloc_tuple (1); Store_field (errorv, 0, Val_int (*error)); args[0] = metacontextv;
(1) "entries" is an API parameter name (of CBArrayAndLen type), so it can be arbitrarily long; pre-patch, it is embedded three times in the same line. Place each occurrence on a separate line. (2) nbd_internal_ocaml_alloc_int64_from_uint32_array() is a humongous function name even for libnbd standards, and even after change (1), it will share a generated line with one instance of the API parameter name. Rename "int64" to "i64" and "uint32" to "u32" in the function's name, to save some characters. While at it, drop a superfluous (harmless) semicolon, dating back to commit 9564dc52abac ("generator: Add OCaml bindings.", 2019-05-30). Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=2172516 Signed-off-by: Laszlo Ersek <lersek@redhat.com> --- generator/OCaml.ml | 9 +++++++-- ocaml/helpers.c | 2 +- ocaml/nbd-c.h | 3 +-- 3 files changed, 9 insertions(+), 5 deletions(-) diff --git a/generator/OCaml.ml b/generator/OCaml.ml index fc29a6a43281..94ab1d589738 100644 --- a/generator/OCaml.ml +++ b/generator/OCaml.ml @@ -524,8 +524,13 @@ let List.iter ( function | CBArrayAndLen (UInt32 n, count) -> - pr " %sv = nbd_internal_ocaml_alloc_int64_from_uint32_array (%s, %s);\n" - n n count; + pr " %sv = " n; + let fncol = output_column () in + let indent = spaces fncol in + pr "nbd_internal_ocaml_alloc_i64_from_u32_array (\n"; + pr "%s %s,\n" indent n; + pr "%s %s\n" indent count; + pr "%s);\n" indent | CBBytesIn (n, len) -> pr " %sv = caml_alloc_initialized_string (%s, %s);\n" n len n | CBInt n | CBUInt n -> diff --git a/ocaml/helpers.c b/ocaml/helpers.c index 8b4b6693a8f1..3361a6968a23 100644 --- a/ocaml/helpers.c +++ b/ocaml/helpers.c @@ -118,7 +118,7 @@ nbd_internal_ocaml_string_list (value ssv) } value -nbd_internal_ocaml_alloc_int64_from_uint32_array (uint32_t *a, size_t len) +nbd_internal_ocaml_alloc_i64_from_u32_array (uint32_t *a, size_t len) { CAMLparam0 (); CAMLlocal2 (v, rv); diff --git a/ocaml/nbd-c.h b/ocaml/nbd-c.h index 8f46415ab16f..f853c84a7756 100644 --- a/ocaml/nbd-c.h +++ b/ocaml/nbd-c.h @@ -61,8 +61,7 @@ extern void nbd_internal_ocaml_raise_error (void) Noreturn; extern void nbd_internal_ocaml_raise_closed (const char *func) Noreturn; extern const char **nbd_internal_ocaml_string_list (value); -extern value nbd_internal_ocaml_alloc_int64_from_uint32_array (uint32_t *, - size_t); +extern value nbd_internal_ocaml_alloc_i64_from_u32_array (uint32_t *, size_t); extern void nbd_internal_unix_sockaddr_to_sa (value, struct sockaddr_storage *, socklen_t *); extern void nbd_internal_ocaml_exception_in_wrapper (const char *, value);
Example effect, in nbd_internal_ocaml_nbd_aio_connect_systemd_socket_activation():
struct nbd_handle *h = NBD_val (hv); if (h == NULL) - nbd_internal_ocaml_raise_closed ("NBD.aio_connect_systemd_socket_activation"); + nbd_internal_ocaml_raise_closed ( + "NBD.aio_connect_systemd_socket_activation" + );
char **argv = (char **) nbd_internal_ocaml_string_list (argvv); int r;
Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=2172516 Signed-off-by: Laszlo Ersek <lersek@redhat.com> --- generator/OCaml.ml | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/generator/OCaml.ml b/generator/OCaml.ml index 94ab1d589738..c49a1f203b9c 100644 --- a/generator/OCaml.ml +++ b/generator/OCaml.ml @@ -633,7 +633,9 @@ let pr " struct nbd_handle *h = NBD_val (hv);\n"; pr " if (h == NULL)\n"; - pr " nbd_internal_ocaml_raise_closed (\"NBD.%s\");\n" name; + pr " nbd_internal_ocaml_raise_closed (\n"; + pr " \"NBD.%s\"\n" name; + pr " );\n"; pr "\n"; List.iter (
Trivially factor out the "num_args" value, for making a subsequent patch easier to review. Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=2172516 Signed-off-by: Laszlo Ersek <lersek@redhat.com> --- generator/OCaml.ml | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/generator/OCaml.ml b/generator/OCaml.ml index c49a1f203b9c..a854afd56171 100644 --- a/generator/OCaml.ml +++ b/generator/OCaml.ml @@ -798,14 +798,15 @@ let pr "}\n"; pr "\n"; - if num_params args optargs > 5 then ( + let num_args = num_params args optargs in + if num_args > 5 then ( pr "/* Byte-code compat function because this method has > 5 parameters.\n"; pr " */\n"; pr "value\n"; pr "nbd_internal_ocaml_nbd_%s_byte (value *argv, int argn)\n" name; pr "{\n"; pr " return nbd_internal_ocaml_nbd_%s (" name; - for i = 0 to num_params args optargs - 1 do + for i = 0 to num_args - 1 do if i > 0 then pr ", "; pr "argv[%d]" i done;
Wrap the long list of (short) arguments in the byte-code compat function bodies that we generate for methods with more than 5 parameters. Effect on "ocaml/nbd-c.c":
@@ -3920,7 +3920,9 @@ value nbd_internal_ocaml_nbd_aio_pread_structured_byte (value *argv, int argn) { - return nbd_internal_ocaml_nbd_aio_pread_structured (argv[0], argv[1], argv[2], argv[3], argv[4], argv[5]); + return nbd_internal_ocaml_nbd_aio_pread_structured (argv[0], argv[1], + argv[2], argv[3], + argv[4], argv[5]); }
value @@ -4242,7 +4244,8 @@ value nbd_internal_ocaml_nbd_aio_block_status_byte (value *argv, int argn) { - return nbd_internal_ocaml_nbd_aio_block_status (argv[0], argv[1], argv[2], argv[3], argv[4], argv[5]); + return nbd_internal_ocaml_nbd_aio_block_status (argv[0], argv[1], argv[2], + argv[3], argv[4], argv[5]); }
value
The patch is easiest to review with "git show -b". Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=2172516 Signed-off-by: Laszlo Ersek <lersek@redhat.com> --- generator/OCaml.ml | 11 +++++++---- 1 file changed, 7 insertions(+), 4 deletions(-) diff --git a/generator/OCaml.ml b/generator/OCaml.ml index a854afd56171..300c8a70d716 100644 --- a/generator/OCaml.ml +++ b/generator/OCaml.ml @@ -806,10 +806,13 @@ let pr "nbd_internal_ocaml_nbd_%s_byte (value *argv, int argn)\n" name; pr "{\n"; pr " return nbd_internal_ocaml_nbd_%s (" name; - for i = 0 to num_args - 1 do - if i > 0 then pr ", "; - pr "argv[%d]" i - done; + let print_args () = + for i = 0 to num_args - 1 do + if i > 0 then pr ", "; + pr "argv[%d]" i + done + in + pr_wrap ',' print_args; pr ");\n"; pr "}\n"; pr "\n"
We're going to reuse the part of "pr_wrap" that runs "code" while temporarily redirecting the output channel to a buffer. Extract it as "pr_buf". Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=2172516 Signed-off-by: Laszlo Ersek <lersek@redhat.com> --- generator/utils.ml | 22 +++++++++++++------- 1 file changed, 14 insertions(+), 8 deletions(-) diff --git a/generator/utils.ml b/generator/utils.ml index 48bd6dd12ba8..5ae87c1ba786 100644 --- a/generator/utils.ml +++ b/generator/utils.ml @@ -151,6 +151,19 @@ let let spaces n = String.make n ' ' +(* Save the current output channel and replace it with a temporary buffer while + * running ‘code’. Return the buffer. + *) +let pr_buf code = + let old_chan = !chan in + let wrapping_col = !col in + let b = Buffer.create 1024 in + chan := Buffer b; + let exn = try code (); None with exn -> Some exn in + chan := old_chan; + col := wrapping_col; + match exn with None -> b | Some exn -> raise exn + (* Wrap the output at maxcol, breaking up lines when a 'c' character * occurs. For example: * foobar = a, b, c, d, e, f, g @@ -165,15 +178,8 @@ let * temporary buffer while running ‘code’. Then we wrap the * buffer and write it to the restored channel. *) - let old_chan = !chan in + let b = pr_buf code in let wrapping_col = !col in - let b = Buffer.create 1024 in - chan := Buffer b; - let exn = try code (); None with exn -> Some exn in - chan := old_chan; - col := wrapping_col; - (match exn with None -> () | Some exn -> raise exn); - let lines = nsplit "\n" (Buffer.contents b) in match lines with | [] -> ()
Explain in "pr_wrap" why the line break is *only seemingly* off-by-one; it is actually correct. Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=2172516 Signed-off-by: Laszlo Ersek <lersek@redhat.com> --- generator/utils.ml | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) diff --git a/generator/utils.ml b/generator/utils.ml index 5ae87c1ba786..919572b73557 100644 --- a/generator/utils.ml +++ b/generator/utils.ml @@ -186,6 +186,9 @@ let | line :: rest -> let fields = nsplit (String.make 1 c) line in let maybe_wrap field = + (* Note that here we break even if we'd fill ‘maxcol‘ precisely; + * that's because... + *) if !col > wrapping_col && !col + String.length field >= maxcol then ( pr "\n%s" (spaces wrapping_col); match span field " \t" with @@ -197,7 +200,11 @@ let let rec loop = function | [] -> () | f :: [] -> let f = maybe_wrap f in pr "%s" f; - | f :: fs -> let f = maybe_wrap f in pr "%s%c" f c; loop fs + | f :: fs -> + let f = maybe_wrap f in + (* ... here we append the separator. *) + pr "%s%c" f c; + loop fs in loop fields;
Add the "pr_wrap_cstr" function, for wrapping C string literals. Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=2172516 Signed-off-by: Laszlo Ersek <lersek@redhat.com> --- generator/utils.mli | 1 + generator/utils.ml | 54 ++++++++++++++++++++ 2 files changed, 55 insertions(+) diff --git a/generator/utils.mli b/generator/utils.mli index 5bbb188fe934..9d6640c87842 100644 --- a/generator/utils.mli +++ b/generator/utils.mli @@ -53,6 +53,7 @@ val val output_to : string -> (unit -> 'a) -> unit val pr : ('a, unit, string, unit) format4 -> 'a val pr_wrap : ?maxcol:int -> char -> (unit -> 'a) -> unit +val pr_wrap_cstr : ?maxcol:int -> (unit -> 'a) -> unit val output_lineno : unit -> int val output_column : unit -> int diff --git a/generator/utils.ml b/generator/utils.ml index 919572b73557..ff1e8a72f7aa 100644 --- a/generator/utils.ml +++ b/generator/utils.ml @@ -213,6 +213,60 @@ let *) pr "%s" (String.concat "\n" rest) +(* Wrap the C string literal output at ‘maxcol’, breaking up lines when a space + * character occurs. For example: + * foobar = "a b c d e f g h i j k" + * └── pr_wrap_cstr ───┘ + * becomes: + * foobar = "a b c d " + * "e f g h " + * "i j k" + * + * Note that: + * - ‘code’ MUST NOT produce the surrounding quotes, + * - ‘code’ MUST NOT produce multiple lines, + * - ‘code’ MUST do its own quoting, + * - space characters produced by ‘code’ cannot be escaped from wrapping. + *) +let pr_wrap_cstr ?(maxcol = 76) code = + (* Just before entering ‘pr_wrap_cstr’, a leading quote must have been + * produced. + *) + let wrapping_col = !col - 1 in + assert (wrapping_col >= 0); + + let b = pr_buf code in + let lines = nsplit "\n" (Buffer.contents b) in + match lines with + | [] -> () + | line :: [] -> + let fields = nsplit " " line in + let nfields = List.length fields in + let indent = spaces wrapping_col in + List.iteri + (fun i field -> + (* Append a space character to each field except the last. *) + let f = if i < nfields - 1 then field ^ " " else field in + + (* Terminate the string literal, insert a line break, and start a + * properly indented new string literal, before printing the field, if + * (a) the field is not the first one in this string literal, and (b) + * printing the field plus a literal-terminating quote would not fit + * in ‘maxcol’. + * + * Note that this way, the literal-terminating quote will always fit + * in ‘maxcol’, except when the *sole* field in the literal is too + * long. + *) + if !col > wrapping_col + 1 && + !col + (String.length f) + 1 > maxcol then + pr "\"\n%s\"" indent; + + (* Print the field. *) + pr "%s" f + ) fields + | _ -> assert false + let output_lineno () = !lineno let output_column () = !col
Going forward, we'd like to wrap long string literals. The recently introduced "pr_wrap_cstr" function may only be called right after the opening quote of the string literal has been produced. In the "print_trace_enter" function, we start a long string literal with "enter:"; however, because we either pass the literal to debug() or debug_direct(), this "enter:" is spelled out twice. Factor "enter:" out to a common line, also ensuring that we'll have as much room for wrapping as possible. Example effect [lib/api.c]:
@@ -6220,7 +6360,8 @@ nbd_supports_uri (struct nbd_handle *h)
/* This function must not call set_error. */ if_debug (h) { - debug_direct (h, "nbd_supports_uri", "enter:"); + debug_direct (h, "nbd_supports_uri", + "enter:"); }
ret = nbd_unlocked_supports_uri (h); @@ -6260,7 +6401,8 @@ nbd_get_uri (struct nbd_handle *h)
pthread_mutex_lock (&h->lock); if_debug (h) { - debug (h, "enter:"); + debug (h, + "enter:"); }
if (unlikely (!get_uri_in_permitted_state (h))) {
Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=2172516 Signed-off-by: Laszlo Ersek <lersek@redhat.com> --- generator/C.ml | 16 +++++++++++----- 1 file changed, 11 insertions(+), 5 deletions(-) diff --git a/generator/C.ml b/generator/C.ml index b65d3de2e82d..0ccf6970d43c 100644 --- a/generator/C.ml +++ b/generator/C.ml @@ -734,11 +734,17 @@ let | Int64 _ | SizeT _ | SockAddrAndLen _ | UInt _ | UInt32 _ | UInt64 _ | UIntPtr _ -> () ) args; - if may_set_error then - pr " debug (h, \"enter:" - else ( - pr " debug_direct (h, \"nbd_%s\", \"enter:" name - ); + let indent = + if may_set_error then ( + pr " debug (h,\n"; + spaces 11 + ) + else ( + pr " debug_direct (h, \"nbd_%s\",\n" name; + spaces 18 + ) in + pr "%s\"" indent; + pr "enter:"; List.iter ( function | Bool n -> pr " %s=%%s" n
On Tue, Apr 25, 2023 at 09:10:55AM +0200, Laszlo Ersek wrote:
Going forward, we'd like to wrap long string literals. The recently introduced "pr_wrap_cstr" function may only be called right after the opening quote of the string literal has been produced. In the "print_trace_enter" function, we start a long string literal with "enter:"; however, because we either pass the literal to debug() or debug_direct(), this "enter:" is spelled out twice. Factor "enter:" out to a common line, also ensuring that we'll have as much room for wrapping as possible.
Example effect [lib/api.c]:
@@ -6220,7 +6360,8 @@ nbd_supports_uri (struct nbd_handle *h)
/* This function must not call set_error. */ if_debug (h) { - debug_direct (h, "nbd_supports_uri", "enter:"); + debug_direct (h, "nbd_supports_uri", + "enter:"); }
ret = nbd_unlocked_supports_uri (h); @@ -6260,7 +6401,8 @@ nbd_get_uri (struct nbd_handle *h)
pthread_mutex_lock (&h->lock); if_debug (h) { - debug (h, "enter:"); + debug (h, + "enter:");
Looks a little funny to not wrap here; but it's generated code, and the context above makes it obvious why it is easier to wrap in both places (when the debug_direct call above has an obvious reason for wrapping). So far, I have no objections to this series, but I've hit the end of my reviewing day. For 1-11: Reviewed-by: Eric Blake <eblake@redhat.com> -- Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3266 Virtualization: qemu.org | libvirt.org
On 4/25/23 23:52, Eric Blake wrote:
On Tue, Apr 25, 2023 at 09:10:55AM +0200, Laszlo Ersek wrote:
Going forward, we'd like to wrap long string literals. The recently introduced "pr_wrap_cstr" function may only be called right after the opening quote of the string literal has been produced. In the "print_trace_enter" function, we start a long string literal with "enter:"; however, because we either pass the literal to debug() or debug_direct(), this "enter:" is spelled out twice. Factor "enter:" out to a common line, also ensuring that we'll have as much room for wrapping as possible.
Example effect [lib/api.c]:
@@ -6220,7 +6360,8 @@ nbd_supports_uri (struct nbd_handle *h)
/* This function must not call set_error. */ if_debug (h) { - debug_direct (h, "nbd_supports_uri", "enter:"); + debug_direct (h, "nbd_supports_uri", + "enter:"); }
ret = nbd_unlocked_supports_uri (h); @@ -6260,7 +6401,8 @@ nbd_get_uri (struct nbd_handle *h)
pthread_mutex_lock (&h->lock); if_debug (h) { - debug (h, "enter:"); + debug (h, + "enter:");
Looks a little funny to not wrap here; but it's generated code, and the context above makes it obvious why it is easier to wrap in both places (when the debug_direct call above has an obvious reason for wrapping). So far, I have no objections to this series, but I've hit the end of my reviewing day.
For 1-11:
Reviewed-by: Eric Blake <eblake@redhat.com>
Thank you! Laszlo
In "print_trace_enter", when we iterate over the "args" and "optargs" lists with "List.iter" for printing the *values*, we use the "function" shorthand within the functions that get applied to each arg / optarg element. Expand the "function" shorthand to "fun" + "match", so we can more easily modify the code subsequently. This patch is purely refactoring, it does not change the generated output. It is best viewed with "git show -b". Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=2172516 Signed-off-by: Laszlo Ersek <lersek@redhat.com> --- generator/C.ml | 42 +++++++++++--------- 1 file changed, 23 insertions(+), 19 deletions(-) diff --git a/generator/C.ml b/generator/C.ml index 0ccf6970d43c..6d880fb953e0 100644 --- a/generator/C.ml +++ b/generator/C.ml @@ -775,27 +775,31 @@ let ) optargs; pr "\""; List.iter ( - function - | Bool n -> pr ", %s ? \"true\" : \"false\"" n - | BytesOut (n, count) - | BytesPersistOut (n, count) -> pr ", %s" count - | BytesIn (n, count) - | BytesPersistIn (n, count) -> - pr ", %s_printable ? %s_printable : \"\", %s" n n count - | Closure { cbname } -> () - | Enum (n, _) -> pr ", %s" n - | Flags (n, _) -> pr ", %s" n - | Fd n | Int n | Int64 n | SizeT n -> pr ", %s" n - | SockAddrAndLen (_, len) -> pr ", (int) %s" len - | Path n | String n | StringList n -> - pr ", %s_printable ? %s_printable : \"\"" n n - | UInt n | UInt32 n | UInt64 n | UIntPtr n -> pr ", %s" n + fun arg -> + (match arg with + | Bool n -> pr ", %s ? \"true\" : \"false\"" n + | BytesOut (n, count) + | BytesPersistOut (n, count) -> pr ", %s" count + | BytesIn (n, count) + | BytesPersistIn (n, count) -> + pr ", %s_printable ? %s_printable : \"\", %s" n n count + | Closure { cbname } -> () + | Enum (n, _) -> pr ", %s" n + | Flags (n, _) -> pr ", %s" n + | Fd n | Int n | Int64 n | SizeT n -> pr ", %s" n + | SockAddrAndLen (_, len) -> pr ", (int) %s" len + | Path n | String n | StringList n -> + pr ", %s_printable ? %s_printable : \"\"" n n + | UInt n | UInt32 n | UInt64 n | UIntPtr n -> pr ", %s" n + ) ) args; List.iter ( - function - | OClosure { cbname } -> - pr ", CALLBACK_IS_NULL (%s_callback) ? \"<fun>\" : \"NULL\"" cbname - | OFlags (n, _, _) -> pr ", %s" n + fun optarg -> + (match optarg with + | OClosure { cbname } -> + pr ", CALLBACK_IS_NULL (%s_callback) ? \"<fun>\" : \"NULL\"" cbname + | OFlags (n, _, _) -> pr ", %s" n + ) ) optargs; pr ");\n"; List.iter (
In "print_trace_enter", for a Closure { "cbname" } arg, we create the format string cbname=<fun> and (accordingly) we don't create a value to print. This is a problem for a subsequent patch. When we iterate over the values to format, it's only the Closure pattern that does not invoke "pr"; all other patterns result in a "pr" call. In order to handle a Closure arg uniformly with the other types, reuse the logic seen with the OClosure optarg pattern. Namely, create the format string cbname=%s and create the value "<fun>". Two examples [lib/api.c]:
@@ -1925,7 +1925,7 @@ nbd_opt_list_meta_context (struct nbd_ha pthread_mutex_lock (&h->lock); if_debug (h) { debug (h, - "enter: context=<fun>"); + "enter: context=%s", "<fun>"); }
if (unlikely (!opt_list_meta_context_in_permitted_state (h))) { @@ -1983,7 +1983,7 @@ nbd_opt_list_meta_context_queries (struc char *queries_printable = nbd_internal_printable_string_list (queries); debug (h, - "enter: queries=%s context=<fun>", queries_printable ? queries_printable : ""); + "enter: queries=%s context=%s", queries_printable ? queries_printable : "", "<fun>"); free (queries_printable); }
Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=2172516 Signed-off-by: Laszlo Ersek <lersek@redhat.com> --- generator/C.ml | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/generator/C.ml b/generator/C.ml index 6d880fb953e0..38b7580b39af 100644 --- a/generator/C.ml +++ b/generator/C.ml @@ -753,7 +753,7 @@ let | BytesIn (n, count) | BytesPersistIn (n, count) -> pr " %s=\\\"%%s\\\" %s=%%zu" n count - | Closure { cbname } -> pr " %s=<fun>" cbname + | Closure { cbname } -> pr " %s=%%s" cbname | Enum (n, _) -> pr " %s=%%d" n | Flags (n, _) -> pr " %s=0x%%x" n | Fd n | Int n -> pr " %s=%%d" n @@ -783,7 +783,7 @@ let | BytesIn (n, count) | BytesPersistIn (n, count) -> pr ", %s_printable ? %s_printable : \"\", %s" n n count - | Closure { cbname } -> () + | Closure _ -> pr ", \"<fun>\"" | Enum (n, _) -> pr ", %s" n | Flags (n, _) -> pr ", %s" n | Fd n | Int n | Int64 n | SizeT n -> pr ", %s" n
The iteration creating the "args" values and the iteration creating the "optargs" values both print commas as prefixes; in other words, whenever any pattern matches, the "pr" that's invoked starts with a ", " substring. For the sake of a subsequent patch, print commas as suffixes, not as prefixes: - Once we're done outputting the format string, calculate the total number of arguments (args + optargs). - If there is at least one argument to print, follow the format string with a comma-space (", "). - Turn both iterations (i.e., over "args" and "optargs") into numbered iterations (i.e., use "List.iteri"). - From each pattern's own "pr" call, strip the comma-space (", ") prefix. - After the "match", see if the (opt)arg just handled was the last one against the *total* number of arguments; if not, print the comma-space as a suffix. This patch is purely refactoring, it does not change the generated output. The command "git show --color-words" is somewhat helpful for viewing the patch. Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=2172516 Signed-off-by: Laszlo Ersek <lersek@redhat.com> --- generator/C.ml | 45 ++++++++++++-------- 1 file changed, 27 insertions(+), 18 deletions(-) diff --git a/generator/C.ml b/generator/C.ml index 38b7580b39af..b4118bf41b3a 100644 --- a/generator/C.ml +++ b/generator/C.ml @@ -774,32 +774,41 @@ let | OFlags (n, _, _) -> pr " %s=0x%%x" n ) optargs; pr "\""; - List.iter ( - fun arg -> + let num_args = List.length args + and num_optargs = List.length optargs in + let num_allargs = num_args + num_optargs in + if num_allargs > 0 then + pr ", "; + List.iteri ( + fun i arg -> (match arg with - | Bool n -> pr ", %s ? \"true\" : \"false\"" n + | Bool n -> pr "%s ? \"true\" : \"false\"" n | BytesOut (n, count) - | BytesPersistOut (n, count) -> pr ", %s" count + | BytesPersistOut (n, count) -> pr "%s" count | BytesIn (n, count) | BytesPersistIn (n, count) -> - pr ", %s_printable ? %s_printable : \"\", %s" n n count - | Closure _ -> pr ", \"<fun>\"" - | Enum (n, _) -> pr ", %s" n - | Flags (n, _) -> pr ", %s" n - | Fd n | Int n | Int64 n | SizeT n -> pr ", %s" n - | SockAddrAndLen (_, len) -> pr ", (int) %s" len + pr "%s_printable ? %s_printable : \"\", %s" n n count + | Closure _ -> pr "\"<fun>\"" + | Enum (n, _) -> pr "%s" n + | Flags (n, _) -> pr "%s" n + | Fd n | Int n | Int64 n | SizeT n -> pr "%s" n + | SockAddrAndLen (_, len) -> pr "(int) %s" len | Path n | String n | StringList n -> - pr ", %s_printable ? %s_printable : \"\"" n n - | UInt n | UInt32 n | UInt64 n | UIntPtr n -> pr ", %s" n - ) + pr "%s_printable ? %s_printable : \"\"" n n + | UInt n | UInt32 n | UInt64 n | UIntPtr n -> pr "%s" n + ); + if i < num_allargs - 1 then + pr ", " ) args; - List.iter ( - fun optarg -> + List.iteri ( + fun i optarg -> (match optarg with | OClosure { cbname } -> - pr ", CALLBACK_IS_NULL (%s_callback) ? \"<fun>\" : \"NULL\"" cbname - | OFlags (n, _, _) -> pr ", %s" n - ) + pr "CALLBACK_IS_NULL (%s_callback) ? \"<fun>\" : \"NULL\"" cbname + | OFlags (n, _, _) -> pr "%s" n + ); + if num_args + i < num_allargs - 1 then + pr ", " ) optargs; pr ");\n"; List.iter (
We're going to want to call "pr_wrap_cstr" in a subsequent patch, for wrapping format strings. "pr_wrap_cstr" only handles a single C string literal, but we may generate a format string that actually consists of multiple string literals. We "exit" and "reenter" string literals only for the various PRI* macros. We can get away with this if we remove the space characters surrounding the PRI* macros; that way, "pr_wrap_cstr" will not be tempted to wrap there -- that is, wrap at a space *outside* of a string literal. Some examples of the effect this patch has [lib/api.c]:
@@ -332,7 +332,7 @@ nbd_set_private_data (struct nbd_handle /* This function must not call set_error. */ if_debug (h) { debug_direct (h, "nbd_set_private_data", - "enter: private_data=%" PRIuPTR "", private_data); + "enter: private_data=%"PRIuPTR"", private_data); }
ret = nbd_unlocked_set_private_data (h, private_data); @@ -2613,7 +2613,7 @@ nbd_connect_vsock (struct nbd_handle *h, pthread_mutex_lock (&h->lock); if_debug (h) { debug (h, - "enter: cid=%" PRIu32 " port=%" PRIu32 "", cid, port); + "enter: cid=%"PRIu32" port=%"PRIu32"", cid, port); }
if (unlikely (!connect_vsock_in_permitted_state (h))) { @@ -3716,7 +3716,7 @@ nbd_pread (struct nbd_handle *h, void *b pthread_mutex_lock (&h->lock); if_debug (h) { debug (h, - "enter: buf=<buf> count=%zu offset=%" PRIu64 " flags=0x%x", count, offset, flags); + "enter: buf=<buf> count=%zu offset=%"PRIu64" flags=0x%x", count, offset, flags); }
if (h->pread_initialize)
Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=2172516 Signed-off-by: Laszlo Ersek <lersek@redhat.com> --- generator/C.ml | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/generator/C.ml b/generator/C.ml index b4118bf41b3a..43601625bcb0 100644 --- a/generator/C.ml +++ b/generator/C.ml @@ -757,16 +757,16 @@ let | Enum (n, _) -> pr " %s=%%d" n | Flags (n, _) -> pr " %s=0x%%x" n | Fd n | Int n -> pr " %s=%%d" n - | Int64 n -> pr " %s=%%\" PRIi64 \"" n + | Int64 n -> pr " %s=%%\"PRIi64\"" n | SizeT n -> pr " %s=%%zu" n | SockAddrAndLen (n, len) -> pr " %s=<sockaddr> %s=%%d" n len | Path n | String n -> pr " %s=%%s" n | StringList n -> pr " %s=%%s" n | UInt n -> pr " %s=%%u" n - | UInt32 n -> pr " %s=%%\" PRIu32 \"" n - | UInt64 n -> pr " %s=%%\" PRIu64 \"" n - | UIntPtr n -> pr " %s=%%\" PRIuPTR \"" n + | UInt32 n -> pr " %s=%%\"PRIu32\"" n + | UInt64 n -> pr " %s=%%\"PRIu64\"" n + | UIntPtr n -> pr " %s=%%\"PRIuPTR\"" n ) args; List.iter ( function
At this point we have ensured that "pr_wrap_cstr" can wrap the "enter" format string, and that "pr_wrap" can wrap the arg/optarg value list. Extract the generation of the format string to "print_format_string" (precisely between the outer quotes), and pass it to "pr_wrap_cstr". If there is at least one argument (arg or optarg alike), don't print a comma-space; after the comma, return the carriage to a new line, and properly indent it, so that it match the name of the debug() vs. debug_direct() function that's being called. This ensures the next "pr_wrap" has maximum width to operate with. This is why we wanted the comma-spaces to be suffixes, rather than prefixes -- we wanted to insert this line break between the first comma after the format string, and the first arg (or optarg, whichever comes first). Extract the generation of the value (arg/optarg) lists to "print_args" (which now only produces inner commas!, see again the prefix->suffix change), and pass "print_args" to "pr_wrap". The longest debug("enter") invocation changes as follows [lib/api.c]:
@@ -5469,7 +5548,10 @@ nbd_aio_pwrite (struct nbd_handle *h, co char *buf_printable = nbd_internal_printable_buffer (buf, count); debug (h, - "enter: buf=\"%s\" count=%zu offset=%"PRIu64" completion=%s flags=0x%x", buf_printable ? buf_printable : "", count, offset, CALLBACK_IS_NULL (completion_callback) ? "<fun>" : "NULL", flags); + "enter: buf=\"%s\" count=%zu offset=%"PRIu64" completion=%s " + "flags=0x%x", + buf_printable ? buf_printable : "", count, offset, + CALLBACK_IS_NULL (completion_callback) ? "<fun>" : "NULL", flags); free (buf_printable); }
The second longest one [lib/api.c]:
@@ -5394,7 +5470,10 @@ nbd_aio_pread_structured (struct nbd_han pthread_mutex_lock (&h->lock); if_debug (h) { debug (h, - "enter: buf=<buf> count=%zu offset=%"PRIu64" chunk=%s completion=%s flags=0x%x", count, offset, "<fun>", CALLBACK_IS_NULL (completion_callback) ? "<fun>" : "NULL", flags); + "enter: buf=<buf> count=%zu offset=%"PRIu64" chunk=%s " + "completion=%s flags=0x%x", + count, offset, "<fun>", + CALLBACK_IS_NULL (completion_callback) ? "<fun>" : "NULL", flags); }
if (h->pread_initialize)
There is one debug_direct("enter") invocation change [lib/api.c]:
@@ -332,7 +335,8 @@ nbd_set_private_data (struct nbd_handle /* This function must not call set_error. */ if_debug (h) { debug_direct (h, "nbd_set_private_data", - "enter: private_data=%"PRIuPTR"", private_data); + "enter: private_data=%"PRIuPTR"", + private_data); }
ret = nbd_unlocked_set_private_data (h, private_data);
This patch is easiest to view with "git show -b". Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=2172516 Signed-off-by: Laszlo Ersek <lersek@redhat.com> --- generator/C.ml | 128 ++++++++++---------- 1 file changed, 67 insertions(+), 61 deletions(-) diff --git a/generator/C.ml b/generator/C.ml index 43601625bcb0..ce6ff69419a5 100644 --- a/generator/C.ml +++ b/generator/C.ml @@ -744,72 +744,78 @@ let spaces 18 ) in pr "%s\"" indent; - pr "enter:"; - List.iter ( - function - | Bool n -> pr " %s=%%s" n - | BytesOut (n, count) - | BytesPersistOut (n, count) -> pr " %s=<buf> %s=%%zu" n count - | BytesIn (n, count) - | BytesPersistIn (n, count) -> - pr " %s=\\\"%%s\\\" %s=%%zu" n count - | Closure { cbname } -> pr " %s=%%s" cbname - | Enum (n, _) -> pr " %s=%%d" n - | Flags (n, _) -> pr " %s=0x%%x" n - | Fd n | Int n -> pr " %s=%%d" n - | Int64 n -> pr " %s=%%\"PRIi64\"" n - | SizeT n -> pr " %s=%%zu" n - | SockAddrAndLen (n, len) -> pr " %s=<sockaddr> %s=%%d" n len - | Path n - | String n -> pr " %s=%%s" n - | StringList n -> pr " %s=%%s" n - | UInt n -> pr " %s=%%u" n - | UInt32 n -> pr " %s=%%\"PRIu32\"" n - | UInt64 n -> pr " %s=%%\"PRIu64\"" n - | UIntPtr n -> pr " %s=%%\"PRIuPTR\"" n - ) args; - List.iter ( - function - | OClosure { cbname } -> pr " %s=%%s" cbname - | OFlags (n, _, _) -> pr " %s=0x%%x" n - ) optargs; + let print_format_string () = + pr "enter:"; + List.iter ( + function + | Bool n -> pr " %s=%%s" n + | BytesOut (n, count) + | BytesPersistOut (n, count) -> pr " %s=<buf> %s=%%zu" n count + | BytesIn (n, count) + | BytesPersistIn (n, count) -> + pr " %s=\\\"%%s\\\" %s=%%zu" n count + | Closure { cbname } -> pr " %s=%%s" cbname + | Enum (n, _) -> pr " %s=%%d" n + | Flags (n, _) -> pr " %s=0x%%x" n + | Fd n | Int n -> pr " %s=%%d" n + | Int64 n -> pr " %s=%%\"PRIi64\"" n + | SizeT n -> pr " %s=%%zu" n + | SockAddrAndLen (n, len) -> pr " %s=<sockaddr> %s=%%d" n len + | Path n + | String n -> pr " %s=%%s" n + | StringList n -> pr " %s=%%s" n + | UInt n -> pr " %s=%%u" n + | UInt32 n -> pr " %s=%%\"PRIu32\"" n + | UInt64 n -> pr " %s=%%\"PRIu64\"" n + | UIntPtr n -> pr " %s=%%\"PRIuPTR\"" n + ) args; + List.iter ( + function + | OClosure { cbname } -> pr " %s=%%s" cbname + | OFlags (n, _, _) -> pr " %s=0x%%x" n + ) optargs + in + pr_wrap_cstr print_format_string; pr "\""; let num_args = List.length args and num_optargs = List.length optargs in let num_allargs = num_args + num_optargs in if num_allargs > 0 then - pr ", "; - List.iteri ( - fun i arg -> - (match arg with - | Bool n -> pr "%s ? \"true\" : \"false\"" n - | BytesOut (n, count) - | BytesPersistOut (n, count) -> pr "%s" count - | BytesIn (n, count) - | BytesPersistIn (n, count) -> - pr "%s_printable ? %s_printable : \"\", %s" n n count - | Closure _ -> pr "\"<fun>\"" - | Enum (n, _) -> pr "%s" n - | Flags (n, _) -> pr "%s" n - | Fd n | Int n | Int64 n | SizeT n -> pr "%s" n - | SockAddrAndLen (_, len) -> pr "(int) %s" len - | Path n | String n | StringList n -> - pr "%s_printable ? %s_printable : \"\"" n n - | UInt n | UInt32 n | UInt64 n | UIntPtr n -> pr "%s" n - ); - if i < num_allargs - 1 then - pr ", " - ) args; - List.iteri ( - fun i optarg -> - (match optarg with - | OClosure { cbname } -> - pr "CALLBACK_IS_NULL (%s_callback) ? \"<fun>\" : \"NULL\"" cbname - | OFlags (n, _, _) -> pr "%s" n - ); - if num_args + i < num_allargs - 1 then - pr ", " - ) optargs; + pr ",\n%s" indent; + let print_args () = + List.iteri ( + fun i arg -> + (match arg with + | Bool n -> pr "%s ? \"true\" : \"false\"" n + | BytesOut (n, count) + | BytesPersistOut (n, count) -> pr "%s" count + | BytesIn (n, count) + | BytesPersistIn (n, count) -> + pr "%s_printable ? %s_printable : \"\", %s" n n count + | Closure _ -> pr "\"<fun>\"" + | Enum (n, _) -> pr "%s" n + | Flags (n, _) -> pr "%s" n + | Fd n | Int n | Int64 n | SizeT n -> pr "%s" n + | SockAddrAndLen (_, len) -> pr "(int) %s" len + | Path n | String n | StringList n -> + pr "%s_printable ? %s_printable : \"\"" n n + | UInt n | UInt32 n | UInt64 n | UIntPtr n -> pr "%s" n + ); + if i < num_allargs - 1 then + pr ", " + ) args; + List.iteri ( + fun i optarg -> + (match optarg with + | OClosure { cbname } -> + pr "CALLBACK_IS_NULL (%s_callback) ? \"<fun>\" : \"NULL\"" cbname + | OFlags (n, _, _) -> pr "%s" n + ); + if num_args + i < num_allargs - 1 then + pr ", " + ) optargs + in + pr_wrap ',' print_args; pr ");\n"; List.iter ( function
Factor out the " ||\n " separator string from the "String.concat" call, and reuse it within the Connected state mapping. An example of the effect [lib/api.c]:
@@ -625,7 +625,8 @@ get_canonical_export_name_in_permitted_s const enum state state = get_public_state (h);
if (!(nbd_internal_is_state_negotiating (state) || - nbd_internal_is_state_ready (state) || nbd_internal_is_state_processing (state) || + nbd_internal_is_state_ready (state) || + nbd_internal_is_state_processing (state) || nbd_internal_is_state_closed (state))) { set_error (nbd_internal_is_state_created (state) ? ENOTCONN : EINVAL, "invalid state: %s: the handle must be %s",
Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=2172516 Signed-off-by: Laszlo Ersek <lersek@redhat.com> --- generator/C.ml | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) diff --git a/generator/C.ml b/generator/C.ml index ce6ff69419a5..f7a306de39d8 100644 --- a/generator/C.ml +++ b/generator/C.ml @@ -533,17 +533,21 @@ let pr "{\n"; pr " const enum state state = get_public_state (h);\n"; pr "\n"; + let or_nl_indent = " ||\n " in let tests = List.map ( function | Created -> "nbd_internal_is_state_created (state)" | Connecting -> "nbd_internal_is_state_connecting (state)" | Negotiating -> "nbd_internal_is_state_negotiating (state)" - | Connected -> "nbd_internal_is_state_ready (state) || nbd_internal_is_state_processing (state)" + | Connected -> + "nbd_internal_is_state_ready (state)" ^ + or_nl_indent ^ + "nbd_internal_is_state_processing (state)" | Closed -> "nbd_internal_is_state_closed (state)" | Dead -> "nbd_internal_is_state_dead (state)" ) permitted_states in - pr " if (!(%s)) {\n" (String.concat " ||\n " tests); + pr " if (!(%s)) {\n" (String.concat or_nl_indent tests); pr " set_error (nbd_internal_is_state_created (state) ? ENOTCONN : EINVAL,\n"; pr " \"invalid state: %%s: the handle must be %%s\",\n"; pr " nbd_internal_state_short_string (state),\n";
The "permitted_state_text" helper function can already fold its output in a certain way, but that's only right for the "generate_docs_nbd_pod" function; i.e., for the "HANDLE STATE" sections of the generated API docs. Therefore, preserve both the folding and non-folding modes of "permitted_state_text", but when we call it (in non-folding mode) from "print_wrapper", apply the previously introduced "pr_wrap_cstr" to the result. The effect on the generated output is [lib/api.c]:
@@ -6522,7 +6522,8 @@ get_uri_in_permitted_state (struct nbd_h set_error (nbd_internal_is_state_created (state) ? ENOTCONN : EINVAL, "invalid state: %s: the handle must be %s", nbd_internal_state_short_string (state), - "connecting, or negotiating, or connected with the server, or shut down, or dead"); + "connecting, or negotiating, or connected with the server, " + "or shut down, or dead"); return false; } return true;
Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=2172516 Signed-off-by: Laszlo Ersek <lersek@redhat.com> --- generator/C.ml | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/generator/C.ml b/generator/C.ml index f7a306de39d8..17b22d84d07d 100644 --- a/generator/C.ml +++ b/generator/C.ml @@ -551,7 +551,9 @@ let pr " set_error (nbd_internal_is_state_created (state) ? ENOTCONN : EINVAL,\n"; pr " \"invalid state: %%s: the handle must be %%s\",\n"; pr " nbd_internal_state_short_string (state),\n"; - pr " \"%s\");\n" (permitted_state_text permitted_states); + pr " \""; + pr_wrap_cstr (fun () -> pr "%s" (permitted_state_text permitted_states)); + pr "\");\n"; pr " return false;\n"; pr " }\n"; pr " return true;\n";
Dependent on "may_set_error", we (a) call debug() vs. debug_direct(), (b) don't pass vs. pass the API name to the function from (a). Currently this is (justifiedly) handled within a single "if"; however, for wrapping the argument lists later -- with differrent indentations --, let's now split the arg lists to separate invocations of "pr", at the cost of another "if may_set_error". This is a refactoring; it does not change the generated output. Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=2172516 Signed-off-by: Laszlo Ersek <lersek@redhat.com> --- generator/C.ml | 11 +++++++---- 1 file changed, 7 insertions(+), 4 deletions(-) diff --git a/generator/C.ml b/generator/C.ml index 17b22d84d07d..7c8fffd89435 100644 --- a/generator/C.ml +++ b/generator/C.ml @@ -854,11 +854,14 @@ let pr " nbd_internal_printable_string (ret);\n" | _ -> () ); - pr " debug (h, \"leave: ret=" + pr " debug (" ) - else ( - pr " debug_direct (h, \"nbd_%s\", \"leave: ret=" name - ); + else + pr " debug_direct ("; + if may_set_error then + pr "h, \"leave: ret=" + else + pr "h, \"nbd_%s\", \"leave: ret=" name; (match ret with | RBool | RErr | RFd | RInt | REnum _ -> pr "%%d\", ret" | RInt64 | RCookie -> pr "%%\" PRIi64, ret"
Similarly to the "print_trace_enter" patch earlier, extract the arg list printing logic to a named function, and pass that one to "pr_wrap". Here we need much less work than in "print_trace_enter": no format string (or other C string literal) can be longer than a line, so we can just wrap all arguments in one go. The effect of this patch is [lib/api.c]:
@@ -247,7 +247,8 @@ nbd_stats_chunks_received (struct nbd_ha ret = nbd_unlocked_stats_chunks_received (h);
if_debug (h) { - debug_direct (h, "nbd_stats_chunks_received", "leave: ret=%" PRIu64, ret); + debug_direct (h, "nbd_stats_chunks_received", "leave: ret=%" PRIu64, + ret); }
if (h->public_state != get_next_state (h)) @@ -1211,7 +1212,8 @@ nbd_get_request_structured_replies (stru ret = nbd_unlocked_get_request_structured_replies (h);
if_debug (h) { - debug_direct (h, "nbd_get_request_structured_replies", "leave: ret=%d", ret); + debug_direct (h, "nbd_get_request_structured_replies", "leave: ret=%d", + ret); }
if (h->public_state != get_next_state (h))
The command "git show -b" is useful with this patch. Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=2172516 Signed-off-by: Laszlo Ersek <lersek@redhat.com> --- generator/C.ml | 31 +++++++++++--------- 1 file changed, 17 insertions(+), 14 deletions(-) diff --git a/generator/C.ml b/generator/C.ml index 7c8fffd89435..cff8f854f1d1 100644 --- a/generator/C.ml +++ b/generator/C.ml @@ -858,20 +858,23 @@ let ) else pr " debug_direct ("; - if may_set_error then - pr "h, \"leave: ret=" - else - pr "h, \"nbd_%s\", \"leave: ret=" name; - (match ret with - | RBool | RErr | RFd | RInt | REnum _ -> pr "%%d\", ret" - | RInt64 | RCookie -> pr "%%\" PRIi64, ret" - | RSizeT -> pr "%%zd\", ret" - | RString -> pr "%%s\", ret_printable ? ret_printable : \"\"" - | RStaticString -> pr "%%s\", ret" - | RUInt | RFlags _ -> pr "%%u\", ret" - | RUIntPtr -> pr "%%\" PRIuPTR, ret" - | RUInt64 -> pr "%%\" PRIu64, ret" - ); + let print_args () = + if may_set_error then + pr "h, \"leave: ret=" + else + pr "h, \"nbd_%s\", \"leave: ret=" name; + (match ret with + | RBool | RErr | RFd | RInt | REnum _ -> pr "%%d\", ret" + | RInt64 | RCookie -> pr "%%\" PRIi64, ret" + | RSizeT -> pr "%%zd\", ret" + | RString -> pr "%%s\", ret_printable ? ret_printable : \"\"" + | RStaticString -> pr "%%s\", ret" + | RUInt | RFlags _ -> pr "%%u\", ret" + | RUIntPtr -> pr "%%\" PRIuPTR, ret" + | RUInt64 -> pr "%%\" PRIu64, ret" + ) + in + pr_wrap ',' print_args; pr ");\n"; if may_set_error then ( (match ret with
Break the flags guard condition, if any, to a new line. One of the most impactful changes of this patch is [lib/api.c]:
@@ -4218,7 +4224,8 @@ nbd_zero (struct nbd_handle *h, uint64_t ret = -1; goto out; } - if (unlikely ((flags & ~0x13) != 0) && ((h->strict & LIBNBD_STRICT_FLAGS) || flags > UINT16_MAX)) { + if (unlikely ((flags & ~0x13) != 0) && + ((h->strict & LIBNBD_STRICT_FLAGS) || flags > UINT16_MAX)) { set_error (EINVAL, "%s: invalid value for flag: 0x%x", "flags", flags); ret = -1;
Note that this patch is not really future-proof; if we add another guard = Some "..."; to "generator/API.ml", that string can still exceed 80 characters. For now, the current patch should suffice. Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=2172516 Signed-off-by: Laszlo Ersek <lersek@redhat.com> --- generator/C.ml | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/generator/C.ml b/generator/C.ml index cff8f854f1d1..cf44424517de 100644 --- a/generator/C.ml +++ b/generator/C.ml @@ -622,7 +622,7 @@ let sprintf "0x%x" !v | None -> "LIBNBD_" ^ flag_prefix ^ "_MASK" in let guard = match guard with - | Some value -> " && " ^ value + | Some value -> " &&\n " ^ value | None -> "" in pr " if (unlikely ((%s & ~%s) != 0)%s) {\n" n mask guard; pr " set_error (EINVAL, \"%%s: invalid value for flag: 0x%%x\",\n";
Use a local boolean flag for unnesting the *_in_permitted_state() function call. The one place where this change currently matters is [lib/api.c]:
@@ -4805,7 +4943,9 @@ nbd_aio_connect_systemd_socket_activatio free (argv_printable); }
- if (unlikely (!aio_connect_systemd_socket_activation_in_permitted_state (h))) { + bool p; + p = aio_connect_systemd_socket_activation_in_permitted_state (h); + if (unlikely (!p)) { ret = -1; goto out; }
Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=2172516 Signed-off-by: Laszlo Ersek <lersek@redhat.com> --- generator/C.ml | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/generator/C.ml b/generator/C.ml index cf44424517de..d5ef067fb119 100644 --- a/generator/C.ml +++ b/generator/C.ml @@ -599,7 +599,9 @@ let let value = match errcode with | Some value -> value | None -> assert false in - pr " if (unlikely (!%s_in_permitted_state (h))) {\n" name; + pr " bool p;\n"; + pr " p = %s_in_permitted_state (h);\n" name; + pr " if (unlikely (!p)) {\n"; pr " ret = %s;\n" value; pr " goto out;\n"; pr " }\n";
On Tue, Apr 25, 2023 at 09:11:06AM +0200, Laszlo Ersek wrote:
Use a local boolean flag for unnesting the *_in_permitted_state() function call.
The one place where this change currently matters is [lib/api.c]:
@@ -4805,7 +4943,9 @@ nbd_aio_connect_systemd_socket_activatio free (argv_printable); }
- if (unlikely (!aio_connect_systemd_socket_activation_in_permitted_state (h))) { + bool p; + p = aio_connect_systemd_socket_activation_in_permitted_state (h);
Declaration after statement. I'm not opposed to it here, but it doesn't seem to be our prevailing style, and can cause issues elsewhere (any function with a goto or switch that jumps past a declaration can cause weird effects for using an uninitialized variable even when the declaration had an initializer; gcc in turn has warnings options that catches those but then flags other late declarations as suspicious). Should we hoist the declaration of p earlier in the generated function? Otherwise, the rest of this series looks good to me. You may want to wait for Rich's review, given the amount of OCaml involved (I may have missed something), but the generated code does look nicer after this series. Feel free to add Reviewed-by: Eric Blake <eblake@rehat.com> throughout the series when pushing. -- Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3266 Virtualization: qemu.org | libvirt.org
On 4/26/23 22:25, Eric Blake wrote:
On Tue, Apr 25, 2023 at 09:11:06AM +0200, Laszlo Ersek wrote:
Use a local boolean flag for unnesting the *_in_permitted_state() function call.
The one place where this change currently matters is [lib/api.c]:
@@ -4805,7 +4943,9 @@ nbd_aio_connect_systemd_socket_activatio free (argv_printable); }
- if (unlikely (!aio_connect_systemd_socket_activation_in_permitted_state (h))) { + bool p; + p = aio_connect_systemd_socket_activation_in_permitted_state (h);
Declaration after statement. I'm not opposed to it here, but it doesn't seem to be our prevailing style, and can cause issues elsewhere (any function with a goto or switch that jumps past a declaration can cause weird effects for using an uninitialized variable even when the declaration had an initializer; gcc in turn has warnings options that catches those but then flags other late declarations as suspicious).
Should we hoist the declaration of p earlier in the generated function?
Yes, will do!
Otherwise, the rest of this series looks good to me. You may want to wait for Rich's review, given the amount of OCaml involved (I may have missed something), but the generated code does look nicer after this series.
Feel free to add Reviewed-by: Eric Blake <eblake@rehat.com> throughout the series when pushing.
Thanks! Laszlo
On 4/26/23 22:25, Eric Blake wrote:
On Tue, Apr 25, 2023 at 09:11:06AM +0200, Laszlo Ersek wrote:
Use a local boolean flag for unnesting the *_in_permitted_state() function call.
The one place where this change currently matters is [lib/api.c]:
@@ -4805,7 +4943,9 @@ nbd_aio_connect_systemd_socket_activatio free (argv_printable); }
- if (unlikely (!aio_connect_systemd_socket_activation_in_permitted_state (h))) { + bool p; + p = aio_connect_systemd_socket_activation_in_permitted_state (h);
Declaration after statement. I'm not opposed to it here, but it doesn't seem to be our prevailing style, and can cause issues elsewhere (any function with a goto or switch that jumps past a declaration can cause weird effects for using an uninitialized variable even when the declaration had an initializer; gcc in turn has warnings options that catches those but then flags other late declarations as suspicious).
Should we hoist the declaration of p earlier in the generated function?
Otherwise, the rest of this series looks good to me. You may want to wait for Rich's review, given the amount of OCaml involved (I may have missed something), but the generated code does look nicer after this series.
Feel free to add Reviewed-by: Eric Blake <eblake@rehat.com> throughout the series when pushing.
I've pushed the first 21 patches as commit range 0744f748ed90..2fa6198f61d4, with your R-bs and Rich's. I'll post a v2 with the last two patches in the series: this patch (v1 #22) needs an update as you're suggesting, and the change effects the *commit message* of the next patch (v1 #23) -- that message includes a diff around nbd_aio_opt_list_meta_context_queries(), and the *context* in that diff is going to change (by my count) once I modify this patch. Thanks! Laszlo
Do the same as commit 0744f748ed90 ("generator: indent C argument list 2 spaces relative to function designator", 2023-04-21), except this time not for the header files, but for "lib/api.c". The most impactful resultant change is [lib/api.c]:
@@ -5378,9 +5577,11 @@ aio_opt_list_meta_context_queries_in_per }
int -nbd_aio_opt_list_meta_context_queries (struct nbd_handle *h, char **queries, - nbd_context_callback context_callback, - nbd_completion_callback completion_callback) +nbd_aio_opt_list_meta_context_queries ( + struct nbd_handle *h, char **queries, + nbd_context_callback context_callback, + nbd_completion_callback completion_callback +) { int ret;
Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=2172516 Signed-off-by: Laszlo Ersek <lersek@redhat.com> --- generator/C.ml | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/generator/C.ml b/generator/C.ml index d5ef067fb119..4f1962832eb0 100644 --- a/generator/C.ml +++ b/generator/C.ml @@ -566,7 +566,8 @@ let let ret_c_type = type_of_ret ret and errcode = errcode_of_ret ret in pr "%s\n" ret_c_type; pr "nbd_%s " name; - print_arg_list ~wrap:true ~handle:true args optargs; + print_arg_list ~wrap:true ~handle:true ~parens:(ParensNewLineWithIndent 0) + args optargs; pr "\n"; pr "{\n"; pr " %s ret;\n" ret_c_type;
All looks good, thanks: Reviewed-by: Richard W.M. Jones <rjones@redhat.com> Rich. -- Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones Read my programming and virtualization blog: http://rwmj.wordpress.com virt-top is 'top' for virtual machines. Tiny program with many powerful monitoring features, net stats, disk stats, logging, etc. http://people.redhat.com/~rjones/virt-top
participants (3)
-
Eric Blake -
Laszlo Ersek -
Richard W.M. Jones