321579 - BeOS build broken creating autoconfig

Status: RESOLVED FIXED

(NSPR :: NSPR, defect)

Description

[Doug Shelton]

User-Agent:       Mozilla/5.0 (BeOS; U; BeOS BePC; en-US; rv:1.9a1) Gecko/20051221 Firefox/1.6a1
Build Identifier: Mozilla/5.0 (BeOS; U; BeOS BePC; en-US; rv:1.9a1) Gecko/20051221 Firefox/1.6a1

BeOS building broke sometime between 2005-12-21 and -26.  Error output is:

make[6]: Entering directory `/boot/home/develop/mozilla/opt-DEBUG/nsprpub/pr/src/linking'
gcc -o prlink.o -c      -pipe -gdwarf-2 -O0   -UNDEBUG -DDEBUG_baron  -DMOZILLA_CLIENT=1 -DDEBUG=1 -DXP_BEOS=1 -DBeOS=1 -DBEOS=1 -D_POSIX_SOURCE=1 -DHAVE_STRERROR=1  -DFORCE_PR_LOG -D_NSPR_BUILD_ -I/boot/home/develop/mozilla/opt-DEBUG/dist/include/nspr -I/boot/home/develop/mozilla/nsprpub/pr/include -I/boot/home/develop/mozilla/nsprpub/pr/include/private  /boot/home/develop/mozilla/nsprpub/pr/src/linking/prlink.c
/boot/home/develop/mozilla/nsprpub/pr/src/linking/prlink.c: In function `PR_GetLibraryFilePathname':
/boot/home/develop/mozilla/nsprpub/pr/src/linking/prlink.c:1393: `Dl_info' undeclared (first use in this function)
/boot/home/develop/mozilla/nsprpub/pr/src/linking/prlink.c:1393: (Each undeclared identifier is reported only once
/boot/home/develop/mozilla/nsprpub/pr/src/linking/prlink.c:1393: for each function it appears in.)
/boot/home/develop/mozilla/nsprpub/pr/src/linking/prlink.c:1393: parse error before `dli'
/boot/home/develop/mozilla/nsprpub/pr/src/linking/prlink.c:1396: `dli' undeclared (first use in this function)
make[6]: *** [prlink.o] Error 1
make[6]: Leaving directory `/boot/home/develop/mozilla/opt-DEBUG/nsprpub/pr/src/linking'
make[5]: *** [export] Error 2
make[5]: Leaving directory `/boot/home/develop/mozilla/opt-DEBUG/nsprpub/pr/src'
make[4]: *** [export] Error 2
make[4]: Leaving directory `/boot/home/develop/mozilla/opt-DEBUG/nsprpub/pr'
make[3]: *** [export] Error 2
make[3]: Leaving directory `/boot/home/develop/mozilla/opt-DEBUG/nsprpub'
make[2]: *** [nspr] Error 2
make[2]: Leaving directory `/boot/home/develop/mozilla/opt-DEBUG'
make[1]: *** [alldep] Error 2
make[1]: Leaving directory `/boot/home/develop/mozilla/opt-DEBUG'
make: *** [alldep] Error 2
$                                  

Reproducible: Always

Steps to Reproduce:
1.attempt to build from CVS
2.
3.

Actual Results:  
build fails

Expected Results:  
build completes

Comment 1

[Sergei Dolgov]

Biesi, isn't it related to
https://bugzilla.mozilla.org/show_bug.cgi?id=321230
?

Comment 2

[Wan-Teh Chang]

Is __GLIBC__ defined on BeOS?

Comment 3

[cls]

Unexpectedly, it is defined. 

/boot/develop/headers/posix/be_setup.h:# define __GLIBC__                      -2

Comment 4

[Wan-Teh Chang]

Is __GLIBC__ defined to be -2? That's a strange value.

Please try changing lines 1391-1392 of
mozilla/nsprpub/pr/src/linking/prlink.c to:

#if defined(USE_DLFCN) && (defined(SOLARIS) || defined(FREEBSD) \
        || defined(LINUX) || defined(__GNU__) || defined(__GLIBC__))

Does that allow the build to complete successfully?

Another possible fix is:

#if defined(SOLARIS) || defined(FREEBSD) \
        || defined(LINUX) || defined(__GNU__) || __GLIBC__ >= 2

Does this also work?

Comment 5

[Christian :Biesinger (don't email me, ping me on IRC)]

(In reply to comment #1)
> https://bugzilla.mozilla.org/show_bug.cgi?id=321230

my build with that patch worked for me. so, I'd say that's unrelated...

Comment 6

[Doug Shelton]

(In reply to comment #4)
> Is __GLIBC__ defined to be -2? That's a strange value.
> 
> Please try changing lines 1391-1392 of
> mozilla/nsprpub/pr/src/linking/prlink.c to:
> 
> #if defined(USE_DLFCN) && (defined(SOLARIS) || defined(FREEBSD) \
>         || defined(LINUX) || defined(__GNU__) || defined(__GLIBC__))
> 
> Does that allow the build to complete successfully?

Yes.  This change allows the build to get past prlink.  Once this build completes, I'll try the second option.

Comment 7

[Doug Shelton]

> 
> Another possible fix is:
> 
> #if defined(SOLARIS) || defined(FREEBSD) \
>         || defined(LINUX) || defined(__GNU__) || __GLIBC__ >= 2
> 
> Does this also work?
> 

No.  With this fix, the build fails as before.

Comment 8

[Wan-Teh Chang]

Could you compile and run this test program on BeOS?
Please compile it with and without "-D_POSIX_SOURCE=1".
Thanks.

#include <stdio.h>

int main()
{
#ifdef __GNU__
    printf("__GNU__ is %d\n", __GNU__);
#endif
#ifdef __GLIBC__
    printf("__GLIBC__ is %d\n", __GLIBC__);
#endif
    return 0;
}

Comment 9

[Doug Shelton]

(In reply to comment #8)
> Could you compile and run this test program on BeOS?
I saved as "glibctest.cpp"  Results are:
$ c++ glibctest.cpp
$ glibctest
__GLIBC__ is 2
$ rm glibctest
$ c++ -D_POSIX_SOURCE=1 glibctest.cpp
$ glibctest
__GLIBC__ is 2
$
 
I hope this helps.

Comment 10

[Sergei Dolgov]

Btw, i think that coming Zeta version will drop glibc and use BSD's version of libc instead. So maybe that definition for POSIX will disappear too.

Comment 11

[Wan-Teh Chang]

Attachment: Proposed patch

This is the fix that Doug Shelton verified in comment 6.
I've checked it in on the NSPRPUB_PRE_4_2_CLIENT_BRANCH
(Mozilla/Gecko 1.9 alpha).

Checking in prlink.c;
/cvsroot/mozilla/nsprpub/pr/src/linking/prlink.c,v  <--  prlink.c
new revision: 3.51.2.30; previous revision: 3.51.2.29
done

Could you nominate someone who is familiar with BeOS's
use of glibc?  I need to ask him to review our patch in
bug 261649, which assumes that glibc is only used by
GNU/Linux, GNU/Hurd, and GNU/k*BSD.

Comment 12

[Robert Millan]

Hi!

Just wanted to clarify a pair of things:

  - My patch didn't really assume Glibc only runs on GNU/Linux, GNU/Hurd, and GNU/k*BSD.  My intention was to assume only that the port of Glibc was complete (I'm not sure if that's the case for BeOS).

  - __GNU__ is defined on GNU/Hurd exclussively, and used everywhere to conditionalise Hurd-specific code, so I don't any non-Hurd system would want to define that macro.

Hope that helps

Comment 13

[Doug Shelton]

(In reply to comment #11)
> Could you nominate someone who is familiar with BeOS's
> use of glibc?  I need to ask him to review our patch in
> bug 261649, which assumes that glibc is only used by
> GNU/Linux, GNU/Hurd, and GNU/k*BSD.
> 
I am not familiar, but I'm hoping fyysik (sergei_d) or biesi might be able to help.

Comment 14

[Doug Shelton]

(In reply to comment #11)
> Could you nominate someone who is familiar with BeOS's
> use of glibc?  I need to ask him to review our patch in
> bug 261649, which assumes that glibc is only used by
> GNU/Linux, GNU/Hurd, and GNU/k*BSD.
> 
The person most familiar is Sergei_d, but he is unavailable for the near future.  I'll inquire elsewhere in the BeOS dev community to see if we can find someone to review this patch.