diff mbox

"Invalid parameter passed to C runtime function" from MinGW GDB

Message ID 838t1qtjty.fsf@gnu.org
State New
Headers show

Commit Message

Eli Zaretskii Nov. 18, 2018, 4:46 p.m. UTC
While debugging the issue with internal_error when stepping out of
'main', I noticed that GDB emits the above warning many times.  It
turns out the warning is triggered by this:

  common/filestuff.c:

  gdb_file_up
  gdb_fopen_cloexec (const char *filename, const char *opentype)
  {
    FILE *result;
    /* Probe for "e" support once.  But, if we can tell the operating
       system doesn't know about close on exec mode "e" without probing,
       skip it.  E.g., the Windows runtime issues an "Invalid parameter
       passed to C runtime function" OutputDebugString warning for
       unknown modes.  Assume that if O_CLOEXEC is zero, then "e" isn't
       supported.  */
    static int fopen_e_ever_failed_einval = O_CLOEXEC == 0;  <<<<<<<<<<<<

    if (!fopen_e_ever_failed_einval)
      {
	char *copy;

	copy = (char *) alloca (strlen (opentype) + 2);
	strcpy (copy, opentype);
	/* This is a glibc extension but we try it unconditionally on
	   this path.  */
	strcat (copy, "e");
	result = fopen (filename, copy);  <<<<<<<<<<<<<<<<<<

MinGW doesn't support the "e" mode.  As you see, we try to be smart
about that, but that trick no longer works, because Gnulib does this:

  build-gnulib/import/fcntl.h:

  #if !defined O_CLOEXEC && defined O_NOINHERIT
  /* Mingw spells it 'O_NOINHERIT'.  */
  # define O_CLOEXEC O_NOINHERIT
  #endif

and O_CLOEXEC is no longer zero.

I propose to fix this as below.  Any objections to committing this to
master (with a suitable ChangeLog, of course)?

Comments

Tom Tromey Nov. 19, 2018, 10:06 p.m. UTC | #1
>>>>> "Eli" == Eli Zaretskii <eliz@gnu.org> writes:

Eli> I propose to fix this as below.  Any objections to committing this to
Eli> master (with a suitable ChangeLog, of course)? 

Eli> diff --git a/gdb/common/filestuff.c b/gdb/common/filestuff.c
Eli> index d4bd1a8..3fa035a 100644
Eli> --- a/gdb/common/filestuff.c
Eli> +++ b/gdb/common/filestuff.c
Eli> @@ -300,8 +300,10 @@ gdb_fopen_cloexec (const char *filename, const char *opentype)
Eli>       skip it.  E.g., the Windows runtime issues an "Invalid parameter
Eli>       passed to C runtime function" OutputDebugString warning for
Eli>       unknown modes.  Assume that if O_CLOEXEC is zero, then "e" isn't
Eli> -     supported.  */
Eli> -  static int fopen_e_ever_failed_einval = O_CLOEXEC == 0;
Eli> +     supported.  On MinGW, O_CLOEXEC is an alias of O_NOINHERIT, and
Eli> +     "e" isn't supported.  */
Eli> +  static int fopen_e_ever_failed_einval =
Eli> +    O_CLOEXEC == 0 || O_CLOEXEC == O_NOINHERIT;

Looks reasonable to me, thanks.

Tom
Eli Zaretskii Nov. 20, 2018, 4:52 p.m. UTC | #2

Eli Zaretskii Nov. 20, 2018, 4:53 p.m. UTC | #3
> From: Tom Tromey <tom@tromey.com>
> Cc: gdb-patches@sourceware.org
> Date: Mon, 19 Nov 2018 15:06:49 -0700
> 
> Eli> diff --git a/gdb/common/filestuff.c b/gdb/common/filestuff.c
> Eli> index d4bd1a8..3fa035a 100644
> Eli> --- a/gdb/common/filestuff.c
> Eli> +++ b/gdb/common/filestuff.c
> Eli> @@ -300,8 +300,10 @@ gdb_fopen_cloexec (const char *filename, const char *opentype)
> Eli>       skip it.  E.g., the Windows runtime issues an "Invalid parameter
> Eli>       passed to C runtime function" OutputDebugString warning for
> Eli>       unknown modes.  Assume that if O_CLOEXEC is zero, then "e" isn't
> Eli> -     supported.  */
> Eli> -  static int fopen_e_ever_failed_einval = O_CLOEXEC == 0;
> Eli> +     supported.  On MinGW, O_CLOEXEC is an alias of O_NOINHERIT, and
> Eli> +     "e" isn't supported.  */
> Eli> +  static int fopen_e_ever_failed_einval =
> Eli> +    O_CLOEXEC == 0 || O_CLOEXEC == O_NOINHERIT;
> 
> Looks reasonable to me, thanks.

Thanks, pushed to master.  (Sorry for an empty message I sent by
mistake a few minutes ago.)
diff mbox

Patch

diff --git a/gdb/common/filestuff.c b/gdb/common/filestuff.c
index d4bd1a8..3fa035a 100644
--- a/gdb/common/filestuff.c
+++ b/gdb/common/filestuff.c
@@ -300,8 +300,10 @@  gdb_fopen_cloexec (const char *filename, const char *opentype)
      skip it.  E.g., the Windows runtime issues an "Invalid parameter
      passed to C runtime function" OutputDebugString warning for
      unknown modes.  Assume that if O_CLOEXEC is zero, then "e" isn't
-     supported.  */
-  static int fopen_e_ever_failed_einval = O_CLOEXEC == 0;
+     supported.  On MinGW, O_CLOEXEC is an alias of O_NOINHERIT, and
+     "e" isn't supported.  */
+  static int fopen_e_ever_failed_einval =
+    O_CLOEXEC == 0 || O_CLOEXEC == O_NOINHERIT;
 
   if (!fopen_e_ever_failed_einval)
     {