Gerald Pfeifer | 1 Feb 01:01 2008

Re: Minor Doc Edits

On Wed, 30 Jan 2008, Tom Browder wrote:
> Following are proposed changes to several gcc/doc/*.texi files for
> "GCC Compiler Collection Internals":

Thanks for this set of patches, Tom!  I tried to extract those that I
felt comfortable with approving myself, updated the ChangeLog entries
as follows, and have committed this after a testing as part of one of
my regular bootstraps.

You'll see that in the case of "CVS removal" I went a step further and 
also avoided two references to the actual system we are using altogether; 
that'll make it easier if one day in a few years we switch to git or 
something else.

In general, somewhat smaller patches usually are processed faster because 
there is less chance that we need more than maintainer to review, or in 
case some issues are clear, and others wait for update/approval/feedback.
I recommend that you resubmit those changes that I couldn't pick up in
at least two parts: one for the "language-specific" change, and one for
the rest.

Also, do you have a copyright assignment with the FSF on file?  Since most
of what I just comitted fixes typos or grammar, that should be below the
limit, but if you expect to contribute more (which I hope ;-), this is a
step we'll need to undertake and I can help with that.

Thanks again for you contribution, and keep the good stuff flowing!

Gerald

(Continue reading)

Ben Elliston | 1 Feb 01:39 2008
Picon

PATCH: fix PR 29326

Andrew Pinski posted a patch for this PR that was discussed on
gcc-patches in November, 2006, but the patch was never committed.  Some
revised text was proposed by Mark Mitchell, which I have used instead.
There was a small grammatical error ("intentionally execution" =>
"intentionally executing"), so I have fixed that.  Tested with `make
info' and inspection.

Okay for mainline?  Any other branches?

Cheers, Ben

2008-02-01  Andrew Pinski  <pinskia <at> gmail.com>
            Mark Mitchell  <mark <at> codesourcery.com>
            Ben Elliston  <bje <at> au.ibm.com>

        PR c/29326
        * doc/extend.texi (Other Builtins): Document.

Index: doc/extend.texi
===================================================================
--- doc/extend.texi     (revision 131994)
+++ doc/extend.texi     (working copy)
 <at>  <at>  -6423,6 +6423,14  <at>  <at>  if (__builtin_expect (ptr != NULL, 1))
 when testing pointer or floating-point values.
  <at> end deftypefn

+ <at> deftypefn {Built-in Function} void __builtin_trap (void)
+This function causes the program to exit abnormally.  GCC implements
+this function by using a target-dependent mechanism (such as
+intentionally executing an illegal instruction) or by calling
(Continue reading)

Marc Gauthier | 1 Feb 02:19 2008

PATCH: Recognizing Xtensa processor variants (1 of 2)

Hi,

This set of patches allows recognizing Xtensa processor variants.
The corresponding change to config.{sub,guess} is already in place.
(The next email only goes out to the binutils list.)

Thanks,
-Marc

2008-01-31  Marc Gauthier  <marc <at> tensilica.com>

	* configure.ac (xtensa*-*-*): Recognize processor variants.
	* configure: Regenerate.

Index: configure.ac
===================================================================
RCS file: /cvs/src/src/configure.ac,v
retrieving revision 1.34
diff -u -p -r1.34 configure.ac
--- configure.ac        31 Jan 2008 10:39:43 -0000      1.34
+++ configure.ac        31 Jan 2008 20:51:39 -0000
 <at>  <at>  -875,7 +875,7  <at>  <at>  case "${target}" in
   vax-*-*)
     noconfigdirs="$noconfigdirs target-newlib target-libgloss ${libgcj}"
     ;;
-  xtensa-*-*)
+  xtensa*-*-*)
     noconfigdirs="$noconfigdirs ${libgcj}"
     ;;
   ip2k-*-*)
(Continue reading)

Mark Mitchell | 1 Feb 02:36 2008

Re: PATCH: fix PR 29326

Ben Elliston wrote:
> Andrew Pinski posted a patch for this PR that was discussed on
> gcc-patches in November, 2006, but the patch was never committed.  Some
> revised text was proposed by Mark Mitchell, which I have used instead.
> There was a small grammatical error ("intentionally execution" =>
> "intentionally executing"), so I have fixed that.  Tested with `make
> info' and inspection.
> 
> Okay for mainline?  Any other branches?

This is OK for mainline, and any other branches open for documentation
changes.

--

-- 
Mark Mitchell
CodeSourcery
mark <at> codesourcery.com
(650) 331-3385 x713

Uros Bizjak | 1 Feb 10:02 2008
Picon

[PATCH, testsuite]: Committed: Update testsuite for --with-arch=core2

Hello!

Attached patch updates the testsuite for newer x86 targets, where we
default to -msse2. The problem is with -mmmx default flag, where some
testcases mix mmx and x87 instructions without emms in between.

2008-02-01  Uros Bizjak  <ubizjak <at> gmail.com>

        PR testsuite/35047
        * gcc.c-torture/execute/20050316-2.x: New file.
        * gcc.target/i386/pr17390.c: Update scan pattern.
        * gcc.dg/builtin-apply4.c: Add -mno-mmx for 32bit x86 targets.
        * g++.dg/tree-ssa/ivopts-1.C: Remove XPASS for x86 targets.

The patch was tested on i686 with -march=core2. Patch is committed to
mainline SVN.

Uros.
Index: gcc.c-torture/execute/20050316-2.x
===================================================================
--- gcc.c-torture/execute/20050316-2.x	(revision 0)
+++ gcc.c-torture/execute/20050316-2.x	(revision 0)
 <at>  <at>  -0,0 +1,9  <at>  <at> 
+# This testcase generates MMX instructions together with x87 instructions.
+# Currently, there is no "emms" generated to switch between register sets,
+# so the testcase fails for targets where MMX insns are enabled.
+
+if { [istarget "i?86-*-*"] || [istarget "x86_64-*-*"] } {
(Continue reading)

Kai Tietz | 1 Feb 12:03 2008

Re: Ping - old patch from April - mingw support for I32/I64 MS printf formatters to c-format.c

"Joseph S. Myers" wrote on 01.02.2008 00:56:40:

> On Wed, 16 Jan 2008, Kai Tietz wrote:
> 
> > ChangeLog of gcc/:
> > 2008-01-15      Kai Tietz  <kai.tietz <at> onevision.com>
> > 
> >         * gcc/c-format.h: Add structure target_ovr_attr to hold
> 
> You need to remove the gcc/ from this and subsequent names in this 
> ChangeLog.
> 
> > +static const char *
> > +replace_formatter_name_to_system_name (const char *attr_name)
> 
> You need a comment above this function explaining what it does 
(including 
> the significance of the parameter and the return value.
> 
> > +  if (attr_name == NULL || *attr_name == 0 || strncmp (attr_name,
> "gcc_", 4) == 0)
> 
> Please try to keep source lines under 80 characters wide, here and 
> elsewhere.
> 
> > +/* Compare the target override format attribute by treating 
> double underscore syntax.  */
> > +static int
> > +compare_tofa (const char *tattr_name, const char *attr_name)
> 
(Continue reading)

Hans-Peter Nilsson | 1 Feb 13:21 2008
Picon

patch ping: support valgrind 3.3

Macro name change with fallback to support < 3.2 and
consolidation of includes and #ifndef/#define chunk.
<http://gcc.gnu.org/ml/gcc-patches/2008-01/msg00890.html>.
Tested on amd64 Debian etch and Fedora 8.

brgds, H-P

Hans-Peter Nilsson | 1 Feb 13:23 2008
Picon

ping doc patch for clz and ctz: add reference to CLZ_DEFINED_VALUE_AT_ZERO and CTZ_DEFINED_VALUE_AT_ZERO

As far as I can tell, the patch is correct.
<http://gcc.gnu.org/ml/gcc-patches/2007-09/msg00992.html>.

brgds, H-P

Toon Moene | 1 Feb 14:21 2008
Picon

Re: [graphite] set CLooG's language to C

Sebastian Pop wrote:

> This is from an email from Tobias Grosser.
> Committed to graphite-branch.
> 
> Sebastian
> 
> The default output language of cloog is FORTRAN. But the FORTRAN output can only
> show complete loops with upper/lower bound, otherwise it fails.

This is not completely true.  The most-used form of a loop construct in 
Fortran certainly is the following:

DO I = 1, N
.... <loop body>
ENDDO

but it is perfectly legal (in Fortran 95, i.e., the language that's 
covered by gfortran) to write:

DO
    IF (condition) EXIT
... <loop body>
ENDDO

or

DO WHILE (condition)
... <loop body>
ENDDO
(Continue reading)

Ismail Dönmez | 1 Feb 14:26 2008
Picon

Re: patch ping: support valgrind 3.3

At Friday 01 February 2008 around 14:21:44 Hans-Peter Nilsson wrote:
> Macro name change with fallback to support < 3.2 and
> consolidation of includes and #ifndef/#define chunk.
> <http://gcc.gnu.org/ml/gcc-patches/2008-01/msg00890.html>.
> Tested on amd64 Debian etch and Fedora 8.

Also tested on i686 on Pardus Linux.

Regards,
ismail

--

-- 
Never learn by your mistakes, if you do you may never dare to try again.


Gmane