[Bug build/27997] New: I get an error when building glibc 2.33.

[Bug build/27997] New: I get an error when building glibc 2.33.

[xer8686 at mail dot ru]

https://sourceware.org/bugzilla/show_bug.cgi?id=27997

            Bug ID: 27997
           Summary: I get an error when building glibc 2.33.
           Product: glibc
           Version: 2.33
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: build
          Assignee: unassigned at sourceware dot org
          Reporter: xer8686 at mail dot ru
                CC: carlos at redhat dot com
  Target Milestone: ---

Created attachment 13504
  --> https://sourceware.org/bugzilla/attachment.cgi?id=13504&action=edit
build logs

I.. -I../libio -I. -nostdinc -isystem
/usr/lib64/gcc/x86_64-openmandriva-linux-gnu/11.1.0/include -isystem
/usr/include -D_LIBC_REENTRANT -include
/builddir/build/BUILD/glibc-2.33/build-i686-linux/libc-modules.h
-DMODULE_NAME=libc -include ../include/libc-symbols.h      
-DTOP_NAMESPACE=glibc -o
/builddir/build/BUILD/glibc-2.33/build-i686-linux/misc/select.o -MD -MP -MF
/builddir/build/BUILD/glibc-2.33/build-i686-linux/misc/select.o.dt -MT
/builddir/build/BUILD/glibc-2.33/build-i686-linux/misc/select.o
../sysdeps/unix/sysv/linux/select.c: In function '__select64':
../sysdeps/unix/sysv/linux/select.c:92:1: error: bp cannot be used in 'asm'
here
   92 | }
      | ^
make[2]: *** [/builddir/build/BUILD/glibc-2.33/build-i686-linux/sysd-rules:213:
/builddir/build/BUILD/glibc-2.33/build-i686-linux/misc/select.o] Error 1
make[2]: Leaving directory '/builddir/build/BUILD/glibc-2.33/misc'
make[1]: *** [Makefile:479: misc/subdir_lib] Error 2
make[1]: Leaving directory '/builddir/build/BUILD/glibc-2.33'
make: *** [Makefile:9: all] Error 2

gcc version 11.1.0

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug build/27997] I get an error when building glibc 2.33.

[xer8686 at mail dot ru]

https://sourceware.org/bugzilla/show_bug.cgi?id=27997

--- Comment #1 from slavaMat <xer8686 at mail dot ru> ---
full log in attachment

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug build/27997] I get an error when building glibc 2.33.

[adhemerval.zanella at linaro dot org]

https://sourceware.org/bugzilla/show_bug.cgi?id=27997

Adhemerval Zanella <adhemerval.zanella at linaro dot org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
           Assignee|unassigned at sourceware dot org   |adhemerval.zanella at linaro dot o
                   |                            |rg
   Last reconfirmed|                            |2021-06-22
                 CC|                            |adhemerval.zanella at linaro dot o
                   |                            |rg
     Ever confirmed|0                           |1

--- Comment #2 from Adhemerval Zanella <adhemerval.zanella at linaro dot org> ---
It is being caused by -mstackrealign where the CAN_USE_REGISTER_ASM_EBP check
at sysdeps/unix/sysv/linux/i386/configure.ac fails to see that -mstackrealign
should set it to 0. 

I need to find a reliable test to check if ebp can be used by the compiler,
regardless of the flags.  Meanwhile you can remove -mstackrealign, gcc default
to 16 bytes aligned stack for a long time.

(there is also a lot of redundant options and some invalid, like the use of
fortification to build glibc itself).

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug build/27997] I get an error when building glibc 2.33.

[fweimer at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=27997

Florian Weimer <fweimer at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |fweimer at redhat dot com

--- Comment #3 from Florian Weimer <fweimer at redhat dot com> ---
-mstackrealign is required to build a glibc which is compatible with legacy
i386 binaries. It cannot be removed easily.

We build with -mstackrealign on i686 and have not encountered this issue.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug build/27997] I get an error when building glibc 2.33.

[adhemerval.zanella at linaro dot org]

https://sourceware.org/bugzilla/show_bug.cgi?id=27997

--- Comment #4 from Adhemerval Zanella <adhemerval.zanella at linaro dot org> ---
The CAN_USE_REGISTER_ASM_EBP tests basically check if -fomit-frame-pointer can
be used on the syscalls generation in C code.  From what I could check the
-fomit-frame-pointer -mstackrealign should set CAN_USE_REGISTER_ASM_EBP to 0,
but the test pass.

I think I have an better idea on how to actually tests it.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug build/27997] I get an error when building glibc 2.33.

[adhemerval.zanella at linaro dot org]

https://sourceware.org/bugzilla/show_bug.cgi?id=27997

--- Comment #5 from Adhemerval Zanella <adhemerval.zanella at linaro dot org> ---
I fact I could reproduce with a simple '-O3 -mstackrealign' flag (but it does
if I add '-march=i686'):

configure CC="x86_64-glibc-linux-gnu-gcc -m32" CFLAGS="-O3 -mstackrealign"
--prefix=/usr

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug build/27997] I get an error when building glibc 2.33.

[xer8686 at mail dot ru]

https://sourceware.org/bugzilla/show_bug.cgi?id=27997

--- Comment #6 from slavaMat <xer8686 at mail dot ru> ---
(In reply to Adhemerval Zanella from comment #2)
> It is being caused by -mstackrealign where the CAN_USE_REGISTER_ASM_EBP
> check at sysdeps/unix/sysv/linux/i386/configure.ac fails to see that
> -mstackrealign should set it to 0. 
> 
> I need to find a reliable test to check if ebp can be used by the compiler,
> regardless of the flags.  Meanwhile you can remove -mstackrealign, gcc
> default to 16 bytes aligned stack for a long time.
> 
> (there is also a lot of redundant options and some invalid, like the use of
> fortification to build glibc itself).

yes when the -mstackrealign flag is removed it builds. Will there be problems
with i386 architectures?

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug build/27997] I get an error when building glibc 2.33.

[fweimer at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=27997

--- Comment #7 from Florian Weimer <fweimer at redhat dot com> ---
(In reply to slavaMat from comment #6)
> yes when the -mstackrealign flag is removed it builds. Will there be
> problems with i386 architectures?

Yes, compatibility with old binaries (before the stack alignment rules changed
to enable SSE2 ) will be reduced; such binaries may crash as a result.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug build/27997] I get an error when building glibc 2.33.

[xer8686 at mail dot ru]

https://sourceware.org/bugzilla/show_bug.cgi?id=27997

--- Comment #8 from slavaMat <xer8686 at mail dot ru> ---
(In reply to Florian Weimer from comment #7)
> (In reply to slavaMat from comment #6)
> > yes when the -mstackrealign flag is removed it builds. Will there be
> > problems with i386 architectures?
> 
> Yes, compatibility with old binaries (before the stack alignment rules
> changed to enable SSE2 ) will be reduced; such binaries may crash as a
> result.

can there be another way than removing the flag?

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug build/27997] I get an error when building glibc 2.33.

[adhemerval.zanella at linaro dot org]

https://sourceware.org/bugzilla/show_bug.cgi?id=27997

--- Comment #9 from Adhemerval Zanella <adhemerval.zanella at linaro dot org> ---
Another option is to set CAN_USE_REGISTER_ASM_EBP to 0 on config.h. I am
working a patch to fix the i386 configure check.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug build/27997] I get an error when building glibc 2.33.

[xer8686 at mail dot ru]

https://sourceware.org/bugzilla/show_bug.cgi?id=27997

--- Comment #10 from slavaMat <xer8686 at mail dot ru> ---
(In reply to Adhemerval Zanella from comment #9)
> Another option is to set CAN_USE_REGISTER_ASM_EBP to 0 on config.h. I am
> working a patch to fix the i386 configure check.

Well, thank you. Please let me know when the fix is.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug build/27997] I get an error when building glibc 2.33.

[xer8686 at mail dot ru]

https://sourceware.org/bugzilla/show_bug.cgi?id=27997

--- Comment #11 from slavaMat <xer8686 at mail dot ru> ---
(In reply to Adhemerval Zanella from comment #9)
> Another option is to set CAN_USE_REGISTER_ASM_EBP to 0 on config.h. I am
> working a patch to fix the i386 configure check.


excuse me you haven't done a patch yet?

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug build/27997] I get an error when building glibc 2.33.

[adhemerval.zanella at linaro dot org]

https://sourceware.org/bugzilla/show_bug.cgi?id=27997

--- Comment #12 from Adhemerval Zanella <adhemerval.zanella at linaro dot org> ---
Nops, the issue seems to be triggered only when some complex code is used along
with some compiler optimization pass. So I couldn't figure out a simple
testcase that reliable show when not to set CAN_USE_REGISTER_ASM_EBP.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug libc/27997] i386: CAN_USE_REGISTER_ASM_EBP is unreliable

[fweimer at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=27997

Florian Weimer <fweimer at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|I get an error when         |i386:
                   |building glibc 2.33.        |CAN_USE_REGISTER_ASM_EBP is
                   |                            |unreliable
                 CC|                            |drepper.fsp at gmail dot com
           See Also|                            |https://gcc.gnu.org/bugzill
                   |                            |a/show_bug.cgi?id=103646
          Component|build                       |libc

--- Comment #13 from Florian Weimer <fweimer at redhat dot com> ---
GCC 12 with vectorization at -O2 makes this more visible.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug libc/27997] i386: CAN_USE_REGISTER_ASM_EBP is unreliable

[fweimer at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=27997

Florian Weimer <fweimer at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |siddhesh at sourceware dot org

--- Comment #14 from Florian Weimer <fweimer at redhat dot com> ---
*** Bug 28680 has been marked as a duplicate of this bug. ***

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug libc/27997] i386: CAN_USE_REGISTER_ASM_EBP is unreliable

[fweimer at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=27997

Florian Weimer <fweimer at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |DUPLICATE
             Status|NEW                         |RESOLVED

--- Comment #15 from Florian Weimer <fweimer at redhat dot com> ---
H.J. requested a new bug for this. It's now fixed in 2.34 and future 2.35.

*** This bug has been marked as a duplicate of bug 28771 ***

-- 
You are receiving this mail because:
You are on the CC list for the bug.