egcs cvs'd between 980214 8:00 and 9:13 UTC on m68k-next-nextstep3:Success

egcs cvs'd between 980214 8:00 and 9:13 UTC on m68k-next-nextstep3:Success

[Toon Moene]

I managed to cvs -z 9 co egcs yesterday 980214 between 9 and 10:13  
local time (UTC + 1).

The subsequent build succeeded without major complaints, although  
of course the new -W -Wall options gave rise to a lot of warnings.

Running c-torture-1.45 showed one test failure to be eliminated  
with respect to the previous snapshot I built (980129):

% diff egcs-980129.ctort egcs-980214.ctort
321,328d320
< ERROR: 961203-1.c: compiler returns exit status 1:
< ERROR: 961203-1.c: compiler returns exit status 1: -O  
-fno-omit-frame-pointer
< ERROR: 961203-1.c: compiler returns exit status 1: -O  
-fomit-frame-pointer
< ERROR: 961203-1.c: compiler returns exit status 1: -O2  
-fno-omit-frame-pointer
< ERROR: 961203-1.c: compiler returns exit status 1: -O2  
-fomit-frame-pointer
< ERROR: 961203-1.c: compiler returns exit status 1: -O3  
-fno-omit-frame-pointer
< ERROR: 961203-1.c: compiler returns exit status 1: -O3  
-fomit-frame-pointer
< ERROR: 961203-1.c: compiler returns exit status 1: -O3  
-funroll-all-loops
561c553
< Test completed with 32 failures
---
> Test completed with 24 failures

Now I want to do the same with the 1.0.2 release branch - what -r  
tag do I specify for that ?

Cheers,
Toon.

Re: egcs cvs'd between 980214 8:00 and 9:13 UTC on m68k-next-nextstep3: Success

[Jeffrey A Law]

  In message < 9802150736.AA06751@moene.indiv.nluug.nl >you write:
  > I managed to cvs -z 9 co egcs yesterday 980214 between 9 and 10:13  
  > local time (UTC + 1).
Cool.  You'll probably find that subsequent updates will run much
faster than the initial checkout :-)


  > The subsequent build succeeded without major complaints, although  
  > of course the new -W -Wall options gave rise to a lot of warnings.
Yup.  I squashed another handful, and, with so many folks fixing
them, I suspect we'll have it down to managable size relatively
quickly.

  > Now I want to do the same with the 1.0.2 release branch - what -r  
  > tag do I specify for that ?
-regcs_1_00_branch

Will check out the 1.0.x branch.
jeff

egcs 1.0.2

[H.J. Lu]

>   > tag do I specify for that ?
> -regcs_1_00_branch
> 
> Will check out the 1.0.x branch.

I have 8 patches for egcs 1.0.2. I have sent 3 of them, which have
yet to be acked. Should I send my other 5 or should I wait for
the backlog to be cleared first?

Thanks.


-- 
H.J. Lu (hjl@gnu.org)

Re: egcs 1.0.2

[Jeffrey A Law]

  In message < m0y46zd-0004ecC@ocean.lucon.org >you write:
  > >   > tag do I specify for that ?
  > > -regcs_1_00_branch
  > > 
  > > Will check out the 1.0.x branch.
  > 
  > I have 8 patches for egcs 1.0.2. I have sent 3 of them, which have
  > yet to be acked. Should I send my other 5 or should I wait for
  > the backlog to be cleared first?
Please wait for the backlog to clear.

Also, are this *critical* patches?  Some of the stuff I've seen go
by look more complex than I would prefer for 1.0.2.

jeff

Re: egcs 1.0.2

[H.J. Lu]

> 
> 
>   In message < m0y46zd-0004ecC@ocean.lucon.org >you write:
>   > >   > tag do I specify for that ?
>   > > -regcs_1_00_branch
>   > > 
>   > > Will check out the 1.0.x branch.
>   > 
>   > I have 8 patches for egcs 1.0.2. I have sent 3 of them, which have
>   > yet to be acked. Should I send my other 5 or should I wait for
>   > the backlog to be cleared first?
> Please wait for the backlog to clear.
> 
> Also, are this *critical* patches?  Some of the stuff I've seen go
> by look more complex than I would prefer for 1.0.2.
> 

One of them is definitely critical. egcs 1.0.1 is broken on x86 at
least with long double. I don't know what else is not working
right. Basically egcs 1.0.1 turns

if (exp1 && exp2)

into

if (exp1)

under certain conditions. As for others, they make it to work on glibc
2.1 and fix some libstdc++ bugs.

BTW, most of them are already in the main trunk. Only 2 patches are
new.

Thanks.

-- 
H.J. Lu (hjl@gnu.org)

Re: egcs 1.0.2

[Jeffrey A Law]

  In message < m0y478b-0004eeC@ocean.lucon.org >you write:
  > One of them is definitely critical. egcs 1.0.1 is broken on x86 at
  > least with long double. I don't know what else is not working
  > right. Basically egcs 1.0.1 turns
I'd hesitate to call a long double bug critical -- few folks use
long double on a regular basis.

jeff

Re: egcs 1.0.2

[H.J. Lu]

> 
> 
>   In message < m0y478b-0004eeC@ocean.lucon.org >you write:
>   > One of them is definitely critical. egcs 1.0.1 is broken on x86 at
>   > least with long double. I don't know what else is not working
>   > right. Basically egcs 1.0.1 turns
> I'd hesitate to call a long double bug critical -- few folks use
> long double on a regular basis.
> 

I cannot use egcs 1.0.x to compile glibc 2 since it uses long double
in many places. That is how I noticed the bug. It is hard to convince
anyone to use egcs 1.0.x if they want to use glibc 2, who happen to
be Debian, RedHat and SUSE.

-- 
H.J. Lu (hjl@gnu.org)

Re: egcs 1.0.2

[Joe Buck]

> I have 8 patches for egcs 1.0.2. I have sent 3 of them, which have
> yet to be acked. Should I send my other 5 or should I wait for
> the backlog to be cleared first?

I gather that there is controversy about spending a lot of effort to
fix "long double" in a minor release.  I would think that if these
changes will not affect code that does not use long double they will
at least do no harm, and not being able to build glibc does make us
look bad.

Re: egcs 1.0.2

[H.J. Lu]

> 
> 
> > I have 8 patches for egcs 1.0.2. I have sent 3 of them, which have
> > yet to be acked. Should I send my other 5 or should I wait for
> > the backlog to be cleared first?
> 
> I gather that there is controversy about spending a lot of effort to
> fix "long double" in a minor release.  I would think that if these

I won't call it "a lot of effort".

> changes will not affect code that does not use long double they will
> at least do no harm, and not being able to build glibc does make us
> look bad.
> 
> 

Here is the patch I have been talking about. It was installed
in egcs evev before egcs 1.0.1 was released.


-- 
H.J. Lu (hjl@gnu.org)
---
Fri Dec  5 16:26:03 1997  Bernd Schmidt <crux@ohara.Informatik.RWTH-Aachen.DE>

	* i386.c (notice_update_cc): Remove bogus pentium GCC code.

--- ./../../../import/egcs/gcc/config/i386/i386.c	Fri Dec 19 00:47:13 1997
+++ config/i386/i386.c	Thu Feb 12 10:49:09 1998
@@ -3397,31 +3397,37 @@
 	  if (cc_status.value1
 	      && reg_overlap_mentioned_p (SET_DEST (exp), cc_status.value1))
 	    cc_status.value1 = 0;
+
 	  if (cc_status.value2
 	      && reg_overlap_mentioned_p (SET_DEST (exp), cc_status.value2))
 	    cc_status.value2 = 0;
+
 	  return;
 	}
+
       /* Moving register into memory doesn't alter the cc's.
 	 It may invalidate the RTX's which we remember the cc's came from.  */
       if (GET_CODE (SET_DEST (exp)) == MEM
 	  && (REG_P (SET_SRC (exp))
 	      || GET_RTX_CLASS (GET_CODE (SET_SRC (exp))) == '<'))
 	{
-	  if (cc_status.value1 && GET_CODE (cc_status.value1) == MEM
-	      || reg_mentioned_p (SET_DEST (exp), cc_status.value1))
+	  if (cc_status.value1
+	      && reg_overlap_mentioned_p (SET_DEST (exp), cc_status.value1))
 	    cc_status.value1 = 0;
-	  if (cc_status.value2 && GET_CODE (cc_status.value2) == MEM
-	      || reg_mentioned_p (SET_DEST (exp), cc_status.value2))
+	  if (cc_status.value2
+	      && reg_overlap_mentioned_p (SET_DEST (exp), cc_status.value2))
 	    cc_status.value2 = 0;
+
 	  return;
 	}
+
       /* Function calls clobber the cc's.  */
       else if (GET_CODE (SET_SRC (exp)) == CALL)
 	{
 	  CC_STATUS_INIT;
 	  return;
 	}
+
       /* Tests and compares set the cc's in predictable ways.  */
       else if (SET_DEST (exp) == cc0_rtx)
 	{
@@ -3429,14 +3435,14 @@
 	  cc_status.value1 = SET_SRC (exp);
 	  return;
 	}
+
       /* Certain instructions effect the condition codes. */
       else if (GET_MODE (SET_SRC (exp)) == SImode
 	       || GET_MODE (SET_SRC (exp)) == HImode
 	       || GET_MODE (SET_SRC (exp)) == QImode)
 	switch (GET_CODE (SET_SRC (exp)))
 	  {
-	  case ASHIFTRT: case LSHIFTRT:
-	  case ASHIFT:
+	  case ASHIFTRT: case LSHIFTRT: case ASHIFT:
 	    /* Shifts on the 386 don't set the condition codes if the
 	       shift count is zero. */
 	    if (GET_CODE (XEXP (SET_SRC (exp), 1)) != CONST_INT)
@@ -3444,6 +3450,7 @@
 		CC_STATUS_INIT;
 		break;
 	      }
+
 	    /* We assume that the CONST_INT is non-zero (this rtx would
 	       have been deleted if it were zero. */
 
@@ -3468,6 +3475,7 @@
       if (SET_DEST (XVECEXP (exp, 0, 0)) == pc_rtx)
 	return;
       if (SET_DEST (XVECEXP (exp, 0, 0)) == cc0_rtx)
+
 	{
 	  CC_STATUS_INIT;
           if (stack_regs_mentioned_p (SET_SRC (XVECEXP (exp, 0, 0))))
@@ -3481,6 +3489,7 @@
 	    cc_status.value1 = SET_SRC (XVECEXP (exp, 0, 0));
 	  return;
 	}
+
       CC_STATUS_INIT;
     }
   else

Re: egcs 1.0.2

[Bernd Schmidt]

> Here is the patch I have been talking about. It was installed
> in egcs evev before egcs 1.0.1 was released.
> 
> Fri Dec  5 16:26:03 1997  Bernd Schmidt <crux@ohara.Informatik.RWTH-Aachen.DE>
> 
> 	* i386.c (notice_update_cc): Remove bogus pentium GCC code.

You are twice wrong: the patch below doesn't match the ChangeLog entry, and
the patch that goes with the ChangeLog entry was installed in egcs-1.0.1.

Bernd

> --- ./../../../import/egcs/gcc/config/i386/i386.c	Fri Dec 19 00:47:13 1997
> +++ config/i386/i386.c	Thu Feb 12 10:49:09 1998
> @@ -3397,31 +3397,37 @@
>  	  if (cc_status.value1
>  	      && reg_overlap_mentioned_p (SET_DEST (exp), cc_status.value1))
>  	    cc_status.value1 = 0;
> +
>  	  if (cc_status.value2
>  	      && reg_overlap_mentioned_p (SET_DEST (exp), cc_status.value2))
>  	    cc_status.value2 = 0;
> +
>  	  return;
>  	}
> +
>        /* Moving register into memory doesn't alter the cc's.
>  	 It may invalidate the RTX's which we remember the cc's came from.  */
>        if (GET_CODE (SET_DEST (exp)) == MEM
>  	  && (REG_P (SET_SRC (exp))
>  	      || GET_RTX_CLASS (GET_CODE (SET_SRC (exp))) == '<'))
>  	{
> -	  if (cc_status.value1 && GET_CODE (cc_status.value1) == MEM
> -	      || reg_mentioned_p (SET_DEST (exp), cc_status.value1))
> +	  if (cc_status.value1
> +	      && reg_overlap_mentioned_p (SET_DEST (exp), cc_status.value1))
>  	    cc_status.value1 = 0;
> -	  if (cc_status.value2 && GET_CODE (cc_status.value2) == MEM
> -	      || reg_mentioned_p (SET_DEST (exp), cc_status.value2))
> +	  if (cc_status.value2
> +	      && reg_overlap_mentioned_p (SET_DEST (exp), cc_status.value2))
>  	    cc_status.value2 = 0;
> +
>  	  return;
>  	}
> +
>        /* Function calls clobber the cc's.  */
>        else if (GET_CODE (SET_SRC (exp)) == CALL)
>  	{
>  	  CC_STATUS_INIT;
>  	  return;
>  	}
> +
>        /* Tests and compares set the cc's in predictable ways.  */
>        else if (SET_DEST (exp) == cc0_rtx)
>  	{
> @@ -3429,14 +3435,14 @@
>  	  cc_status.value1 = SET_SRC (exp);
>  	  return;
>  	}
> +
>        /* Certain instructions effect the condition codes. */
>        else if (GET_MODE (SET_SRC (exp)) == SImode
>  	       || GET_MODE (SET_SRC (exp)) == HImode
>  	       || GET_MODE (SET_SRC (exp)) == QImode)
>  	switch (GET_CODE (SET_SRC (exp)))
>  	  {
> -	  case ASHIFTRT: case LSHIFTRT:
> -	  case ASHIFT:
> +	  case ASHIFTRT: case LSHIFTRT: case ASHIFT:
>  	    /* Shifts on the 386 don't set the condition codes if the
>  	       shift count is zero. */
>  	    if (GET_CODE (XEXP (SET_SRC (exp), 1)) != CONST_INT)
> @@ -3444,6 +3450,7 @@
>  		CC_STATUS_INIT;
>  		break;
>  	      }
> +
>  	    /* We assume that the CONST_INT is non-zero (this rtx would
>  	       have been deleted if it were zero. */
>  
> @@ -3468,6 +3475,7 @@
>        if (SET_DEST (XVECEXP (exp, 0, 0)) == pc_rtx)
>  	return;
>        if (SET_DEST (XVECEXP (exp, 0, 0)) == cc0_rtx)
> +
>  	{
>  	  CC_STATUS_INIT;
>            if (stack_regs_mentioned_p (SET_SRC (XVECEXP (exp, 0, 0))))
> @@ -3481,6 +3489,7 @@
>  	    cc_status.value1 = SET_SRC (XVECEXP (exp, 0, 0));
>  	  return;
>  	}
> +
>        CC_STATUS_INIT;
>      }
>    else
> 

Re: egcs 1.0.2

[H.J. Lu]

> 
> > Here is the patch I have been talking about. It was installed
> > in egcs evev before egcs 1.0.1 was released.
> > 
> > Fri Dec  5 16:26:03 1997  Bernd Schmidt <crux@ohara.Informatik.RWTH-Aachen.DE>
> > 
> > 	* i386.c (notice_update_cc): Remove bogus pentium GCC code.
> 
> You are twice wrong: the patch below doesn't match the ChangeLog entry, and
> the patch that goes with the ChangeLog entry was installed in egcs-1.0.1.
> 

I was wrong. Sorry for that. I was very frustrated since egcs 1.0.1
miscocmpiled glibc 2. It appears that the patch really came from

Mon Sep 29 08:21:35 1997  Bruno Haible <bruno@linuix.mathematik.uni-karlsruhe.de>

	* i386.c (notice_update_cc): Use reg_overlap_mentioned_p.

which is not in egcs 1.0.1.

Thanks.

H.J.

Re: egcs 1.0.2

[Bruno Haible]

H.J.:
> It appears that the patch really came from
>
> Mon Sep 29 08:21:35 1997  Bruno Haible 
> 
>         * i386.c (notice_update_cc): Use reg_overlap_mentioned_p.
> 
> which is not in egcs 1.0.1.

I apologize for having sent that patch to Kenner only, not to this list.
Here it is again.


egcs-1.0.1 has a bug which wasn't present in gcc-2.7.2.
A "testl $64,%eax" or "testb $64,-4(%ebp)" is omitted. This kind of bug
occurs in particular when DImode variables are used. Please find below
a problem report and a fix.

========================= foo.c ======================
typedef struct { int addr; int type; } object;

extern object bar (object);
object foo (object x, object y)
{
  object z = *(object*)(x.addr);
  if (z.type & 64)
    {
      y = *(object*)(z.addr+8);
      z = *(object*)(z.addr);
      if (z.type & 64)
        y = bar(y);
    }
  return y;
}
======================================================
$ ./xgcc -B./ -O -S foo.c && cat foo.s
	.file	"foo.c"
	.version	"01.01"
/ GNU C version egcs-2.90.23 980102 (egcs-1.0.1 release) (i586-pc-linux-gnulibc1) compiled by GNU C version 2.7.2.
/ options passed:  -O
/ options enabled:  -fdefer-pop -fthread-jumps -fpeephole -ffunction-cse
/ -finline -fkeep-static-consts -fpcc-struct-return -fcommon -fverbose-asm
/ -fgnu-linker -falias-check -fargument-alias -m80387 -mhard-float
/ -mno-soft-float -mieee-fp -mfp-ret-in-387 -mschedule-prologue
/ -mcpu=pentium -march=pentium

gcc2_compiled.:
.text
	.align 4
.globl foo
	.type	 foo,@function
foo:
	pushl %ebp
	movl %esp,%ebp
	subl $8,%esp
	pushl %esi
	pushl %ebx
	movl 8(%ebp),%ebx
	movl 12(%ebp),%eax
	movl (%eax),%edx
	movl %edx,-8(%ebp)
	movl 4(%eax),%eax
	movl %eax,-4(%ebp)
	testb $64,-4(%ebp)
	je .L2
	movl 8(%edx),%esi
	movl %esi,20(%ebp)
	movl 12(%edx),%ecx
	movl %ecx,24(%ebp)
	movl (%edx),%eax
	movl %eax,-8(%ebp)
	movl 4(%edx),%eax
	movl %eax,-4(%ebp)
				;; <========== "testl $64,%eax" missing here
	je .L2
	leal 20(%ebp),%eax
	pushl %ecx
	pushl %esi
	pushl %eax
	call bar
.L2:
	movl 20(%ebp),%eax
	movl %eax,(%ebx)
	movl 24(%ebp),%eax
	movl %eax,4(%ebx)
	movl %ebx,%eax
	leal -16(%ebp),%esp
	popl %ebx
	popl %esi
	movl %ebp,%esp
	popl %ebp
	ret $4
.Lfe1:
	.size	 foo,.Lfe1-foo
	.ident	"GCC: (GNU) egcs-2.90.23 980102 (egcs-1.0.1 release)"
===============================================================================

Here is the fix, which is already present in gcc-2.8.0. (Incidentally,
Kenner mutilated my changelog entry when putting in the patch.)

Sun Sep 28 03:45:32 1997  Bruno Haible  <bruno@linuix.mathematik.uni-karlsruhe.de>

        * i386/i386.c (notice_update_cc): Use reg_overlap_mentioned_p, not
        reg_mentioned_p, for cc_status.value1 might be a ZERO_EXTRACT rtx.

===============================================================================
*** egcs-1.0.1/gcc/config/i386/i386.c.bak	Fri Dec 19 09:47:13 1997
--- egcs-1.0.1/gcc/config/i386/i386.c	Mon Feb 16 22:59:43 1998
***************
*** 3408,3418 ****
  	  && (REG_P (SET_SRC (exp))
  	      || GET_RTX_CLASS (GET_CODE (SET_SRC (exp))) == '<'))
  	{
! 	  if (cc_status.value1 && GET_CODE (cc_status.value1) == MEM
! 	      || reg_mentioned_p (SET_DEST (exp), cc_status.value1))
  	    cc_status.value1 = 0;
! 	  if (cc_status.value2 && GET_CODE (cc_status.value2) == MEM
! 	      || reg_mentioned_p (SET_DEST (exp), cc_status.value2))
  	    cc_status.value2 = 0;
  	  return;
  	}
--- 3408,3418 ----
  	  && (REG_P (SET_SRC (exp))
  	      || GET_RTX_CLASS (GET_CODE (SET_SRC (exp))) == '<'))
  	{
! 	  if (cc_status.value1
! 	      && reg_overlap_mentioned_p (SET_DEST (exp), cc_status.value1))
  	    cc_status.value1 = 0;
! 	  if (cc_status.value2
! 	      && reg_overlap_mentioned_p (SET_DEST (exp), cc_status.value2))
  	    cc_status.value2 = 0;
  	  return;
  	}
===============================================================================

Can you please put in this patch in the devel branch and later into
egcs-1.0.2 ?

Bruno