Re: [CFT] sed prerelease on alpha.gnu.org

[Matthew Woehlke]

Paolo Bonzini wrote:
> Matthew Woehlke wrote:
> > Paolo Bonzini wrote:
> > > I placed a prerelease on ftp://alpha.gnu.org/gnu/sed/sed-4.1e.tar.gz
> > > and I would like as many people as possible to test it on various
> > > architectures and operating systems.  This is the first fully
> > > gnulib-ized version of sed.
> > It seems to build rather more reliably than 4.1.5 :-).
>
> Can you try sed 4.1f too?  I fixed the AIX failure and tried XFAILing
> the utf8-* tests where they do not work.  

AIX:

source='compile.c' object='compile.o' libtool=no \
        DEPDIR=.deps depmode=aix /bin/sh ../build-aux/depcomp \
        cc -qlanglvl=ansi -DHAVE_CONFIG_H -I. -I..  -I../lib -I.. 
-I../lib -DLOCALEDIR=\"/home/install/gnu/rs6000_aix/share/locale\" 
-I/home/install/gnu/rs6000_aix/include 
-I/home/install/gnu/rs6000_aix/include -c compile.c
"compile.c", line 532.1: 1506-343 (S) Redeclaration of match_slash 
differs from previous declaration on line 530 of "compile.c".
"compile.c", line 532.1: 1506-379 (I) Prototype for function match_slash 
must contain only promoted types if prototype and nonprototype 
declarations are mixed.
"compile.c", line 532.1: 1506-380 (I) Parameter 2 has type "signed char" 
which promotes to "int".

Okay... the problem here is that "[prototypes for functions] must 
contain only promoted types if prototype and nonprototype declarations 
are mixed." The c89-definition of "bool" is biting you here, and will 
bite you wherever 'bool' appears in a function prototype.

For now I'm just going to use 'CC="cc -qlanglvl=stdc99"'. I don't know 
how hard you want to try to make this work out of the box, but the 
previous change should probably be reverted.

FYI: adding the autoconf check to turn on c99 mode is broken; autoconf 
picks up -qlanglvl=ansi from the ANSI c89 check, and doesn't know how to 
try alternatives (this affects other GNU packages as well, most notably 
coreutils).

> The XFAILing should work on
> HP-UX and Windows; I do not have access to Solaris.

the XFAIL'ing is apparently working on Solaris also. No change on Irix 
(not that any was expected).

> > Successful build on x86/FreeBSD. Seems okay besides the noted utf8-*
> > tests on sparc/Solaris, x86/Solaris, ia64/HP-UX, Irix (the last of which
> > skipped said tests).
>
> Skipping is okay. :-)

Of course :-). I'm just reporting my results.

> > However risc/HP-UX went poorly, most tests produced
> > the failure (e.g.) "../sed/sed: file ./numsub.sed line 3: Memory
> > exhausted" (though this seems like it may be a problem with the
> > machine...).
>
> It may also be a problem with the system malloc, for example returning
> ENOMEM for a zero-byte request.  If you are familiar with gnulib, [snip]

Not really, sorry. (Maybe if I have time later I'll go digging around 
for source checkouts and such; right now I should go do some $REALWORK...)

However, while I want to say at least one of the platforms I'm building 
on has such a malloc...

$ cat malloc-test.c
#include <stdio.h>
#include <stdlib.h>

int main() {
  errno = 0;
  void *foo = malloc(0);
  printf("ptr = %p, errno = %i\n", foo, errno);
  return 0;
}
$ cc -o mt malloc-test.c && ./mt
ptr = 40003160, errno = 0
$ uname -svriomp
HP-UX B.11.11 U 9000/800 unknown unknown HP-UX

--
Matthew
Please do not quote my e-mail address unobfuscated in message bodies.
--
This is not a sig. I am too lazy to steal one, perhaps you could loan me 
yours? -- Unknown