All of lore.kernel.org
 help / color / mirror / Atom feed
From: Cyril Hrubis <chrubis@suse.cz>
To: ltp@lists.linux.it
Subject: [LTP] [RFC] [PATCH] lib32: Add support for linking 32bit programs	on 64bit
Date: Wed, 29 Mar 2017 11:12:50 +0200	[thread overview]
Message-ID: <20170329091250.GC7089@rei.lan> (raw)
In-Reply-To: <1888940677.16511157.1490776585392.JavaMail.zimbra@redhat.com>

Hi!
> > * Configure test if compiler supports -m32
> >   TODO: Does 32bit compiler accept -m32?
> >         If so, we should fail the test on 32bit
> > 	platforms by default.
> > 
> > * The directory lib32 contains symlinks of minimal set of sources
> >   needed to link newlib testcase + Makefile.
> > 
> > * The testcases.mk builds 32bit library as well if
> >   32bit compiler was installed (-m32 configure test was succesful)
> > 
> > * And finally the env_post.mk switches between lib and lib32 depeding
> >   on if -m32 was passed in the CFLAGS
> > 
> 
> One issue I ran into is that configure checks can vary
> between default and -m32. Here's one exception from RHEL5.6:
> 
> # cat 1.c 
> int main(void) {
>         int i = 0;
>         return __sync_add_and_fetch(&i, 1);
> }
> 
> # gcc 1.c 
> # gcc -m32 1.c 
> /tmp/ccwrlcJi.o: In function `main':
> 1.c:(.text+0x27): undefined reference to `__sync_add_and_fetch_4'
> collect2: ld returned 1 exit status

Right, we would have to check for atomic operations with -m32 as well.

I guess that we can add -DUSE_M32 into the lib32/Makefile and add a few
ifdefs into the tst_atomic.h something as:

#if (defined(USE_M32) && HAVE_SYNC_AND_FETCH32 == 1) || \
    (!defined(USE_M32) && HAVE_SYNC_AND_FETCH == 1)
...
#elif ...
...

> > TODO: Check if LTP works with -m32 in CFLAGS passed to configure as well
> 
> This fails for me on RHEL7:
> 
> # make
> make[1]: Entering directory `/root/ltp/lib/newlib_tests'
> gcc -m32 -g -O2 -fno-strict-aliasing -pipe -Wall -W -Wold-style-definition -W -Wall  -I../../include -I../../include -I../../include/old/   -L../../lib32  test14.c   -lltp -o test14
> /tmp/ccPZUDSr.o: In function `cleanup':
> /root/ltp/lib/newlib_tests/test14.c:31: undefined reference to `safe_fopen'
> /root/ltp/lib/newlib_tests/test14.c:32: undefined reference to `safe_socket'
> collect2: error: ld returned 1 exit status
> make[1]: *** [test14] Error 1
> make[1]: Leaving directory `/root/ltp/lib/newlib_tests'
> make: *** [all] Error 2

This is likely because we try to link against lib32 since -m32 is passed
in CFLAGS but lib32/ directory links only subset of the C sources. I
would like to disable all the lib32 machinery either when we run on
32bit machine or if -m32 was passed to the configure and use lib32 only
in a case that we run on 64 bit machine and want to link 32 bit binary.

-- 
Cyril Hrubis
chrubis@suse.cz

  reply	other threads:[~2017-03-29  9:12 UTC|newest]

Thread overview: 9+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2017-03-28 10:01 [LTP] [RFC] [PATCH] lib32: Add support for linking 32bit programs on 64bit Cyril Hrubis
2017-03-28 14:55 ` Richard Palethorpe
2017-03-28 15:33   ` Cyril Hrubis
2017-03-28 15:46     ` Cyril Hrubis
2017-03-29  9:37       ` Richard Palethorpe
2017-03-29  8:36 ` Jan Stancek
2017-03-29  9:12   ` Cyril Hrubis [this message]
2017-03-29 12:48 ` Richard Palethorpe
2017-03-29 14:56   ` Richard Palethorpe

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20170329091250.GC7089@rei.lan \
    --to=chrubis@suse.cz \
    --cc=ltp@lists.linux.it \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.