[basic-improvements] Merge from trunk

[basic-improvements] Merge from trunk

[Zack Weinberg]

The merge tag is gcc-3_4-basic-improvements-merge-20021105.
Bootstrapped i686-linux; there are regressions relative to mainline,
but they are not caused by this merge.

--- ws.main     2002-11-05 11:10:55.000000000 -0800
+++ ws.bib      2002-11-05 11:10:41.000000000 -0800
@@ -28,9 +28,12 @@
 FAIL: gcc.c-torture/execute/builtin-constant.c execution,  -O3 -fomit-frame-pointer 
 FAIL: gcc.c-torture/execute/builtin-constant.c execution,  -O3 -g 
 FAIL: gcc.c-torture/execute/builtin-constant.c execution,  -Os 
+FAIL: gcc.dg/i386-sse-1.c (test for excess errors)
+FAIL: gcc.dg/i386-sse-2.c (test for excess errors)
+FAIL: gcc.dg/i386-sse-3.c (test for excess errors)
                === gcc Summary ===
-# of expected passes           20539
-# of unexpected failures       5
+# of expected passes           20552
+# of unexpected failures       8
 # of expected failures         64
 # of unsupported tests         65
                === g++ tests ===

I believe I know what is causing these failures, and will send a
separate patch for them shortly.

I would appreciate it if the HPPA maintainers would check the merge of
pa.c; the function-call rewrite on the mainline clashed a bit with the
conversion of ASM_OUTPUT_INTERNAL_LABEL to a target hook on the branch.
I believe I got it correct but I don't have access to this platform to
test it.

zw

Re: [basic-improvements] Merge from trunk

[Jan Hubicka]

> The merge tag is gcc-3_4-basic-improvements-merge-20021105.
> Bootstrapped i686-linux; there are regressions relative to mainline,
> but they are not caused by this merge.
> 
> --- ws.main     2002-11-05 11:10:55.000000000 -0800
> +++ ws.bib      2002-11-05 11:10:41.000000000 -0800
> @@ -28,9 +28,12 @@
>  FAIL: gcc.c-torture/execute/builtin-constant.c execution,  -O3 -fomit-frame-pointer 
>  FAIL: gcc.c-torture/execute/builtin-constant.c execution,  -O3 -g 
>  FAIL: gcc.c-torture/execute/builtin-constant.c execution,  -Os 
> +FAIL: gcc.dg/i386-sse-1.c (test for excess errors)
> +FAIL: gcc.dg/i386-sse-2.c (test for excess errors)
> +FAIL: gcc.dg/i386-sse-3.c (test for excess errors)
>                 === gcc Summary ===
> -# of expected passes           20539
> -# of unexpected failures       5
> +# of expected passes           20552
> +# of unexpected failures       8
>  # of expected failures         64
>  # of unsupported tests         65
>                 === g++ tests ===
> 
> I believe I know what is causing these failures, and will send a
> separate patch for them shortly.
If you will run into problem, just let me know and I will check it.
After all I probably caused it.


Thanks!
Honza
> 
> I would appreciate it if the HPPA maintainers would check the merge of
> pa.c; the function-call rewrite on the mainline clashed a bit with the
> conversion of ASM_OUTPUT_INTERNAL_LABEL to a target hook on the branch.
> I believe I got it correct but I don't have access to this platform to
> test it.
> 
> zw

[basic-improvements] Merge from trunk

[Zack Weinberg]

The merge tag is gcc-3_4-basic-improvements-merge-20021202.
Bootstrapped i686-linux with no regressions relative to mainline.

zw

[basic-improvements] Merge from trunk

[Zack Weinberg]

The merge tag is gcc-3_4-basic-improvements-merge-20021027.
Bootstrapped i686-linux without Ada - Ada is broken on mainline at
present (the Storage_Error ICE reported elsewhere).

Both mainline and branch are crashing on a variety of test cases at
-O1 and only -O1.  Failures are of the form

20020604-1.c: In function `foo':
20020604-1.c:87: internal compiler error: virtual array reg_n_info[92]: 
 element 92 out of bounds in init_alias_analysis, at alias.c:2814

or

920828-1.c: In function `f':
920828-1.c:1: internal compiler error: Segmentation fault

I may look into this later today, but no promises.

zw

Re: [basic-improvements] Merge from trunk

[Andreas Jaeger]

Zack Weinberg <zack@codesourcery.com> writes:

> The merge tag is gcc-3_4-basic-improvements-merge-20021027.
> Bootstrapped i686-linux without Ada - Ada is broken on mainline at
> present (the Storage_Error ICE reported elsewhere).
>
> Both mainline and branch are crashing on a variety of test cases at
> -O1 and only -O1.  Failures are of the form
>
> 20020604-1.c: In function `foo':
> 20020604-1.c:87: internal compiler error: virtual array reg_n_info[92]: 
>  element 92 out of bounds in init_alias_analysis, at alias.c:2814

This should be the same problem as the Ada problem, I suggest to test
Honza's patch before spending any more time on it.  Hopefully soon
somebody finds time to review Honza's patch for that problem...

> or
>
> 920828-1.c: In function `f':
> 920828-1.c:1: internal compiler error: Segmentation fault
>
> I may look into this later today, but no promises.
>
> zw
>

Andreas
-- 
 Andreas Jaeger
  SuSE Labs aj@suse.de
   private aj@arthur.inka.de
    http://www.suse.de/~aj

Re: [basic-improvements] Merge from trunk

[Zack Weinberg]

On Mon, Oct 28, 2002 at 08:44:12PM +0100, Andreas Jaeger wrote:
> Zack Weinberg <zack@codesourcery.com> writes:
> 
> > The merge tag is gcc-3_4-basic-improvements-merge-20021027.
> > Bootstrapped i686-linux without Ada - Ada is broken on mainline at
> > present (the Storage_Error ICE reported elsewhere).
> >
> > Both mainline and branch are crashing on a variety of test cases at
> > -O1 and only -O1.  Failures are of the form
> >
> > 20020604-1.c: In function `foo':
> > 20020604-1.c:87: internal compiler error: virtual array reg_n_info[92]: 
> >  element 92 out of bounds in init_alias_analysis, at alias.c:2814
> 
> This should be the same problem as the Ada problem, I suggest to test
> Honza's patch before spending any more time on it.  Hopefully soon
> somebody finds time to review Honza's patch for that problem...

Yes, that patch makes all the check-gcc failures that I was seeing go
away.  Thanks for the pointer.

zw

[basic-improvements] Merge from trunk

[Zack Weinberg]

The merge tag is gcc-3_4-basic-improvements-merge-20021020.
Bootstrapped i686-linux, all languages including Ada, with no
regressions relative to the merge point.

$ diff -u LOG.trunk LOG.bib
--- LOG.trunk  2002-10-21 10:41:54 -0700
+++ LOG.bib    2002-10-21 10:40:51 -0700
@@ -17,9 +17,10 @@
                === gcc tests ===
 FAIL: gcc.dg/noncompile/incomplete-1.c (test for excess errors)
                === gcc Summary ===
-# of expected passes           20752
+# of expected passes           20766
 # of unexpected failures       1
 # of expected failures         64
+# of unresolved testcases      1
 # of unsupported tests         65
                === g++ tests ===
 FAIL: g++.dg/inherit/override1.C (test for excess errors)

The additional unresolved comes from a typo in a test case only
present on the branch.  I will be sending a separate patch for that
issue shortly.

zw