Recent changes to regex.c break bootstrapping

Recent changes to regex.c break bootstrapping

[Luc Teirlinck]

After bootstrapping failed due to an undefined reference to xalloc, I
made the obvious correction to xmalloc and installed in CVS, but after
that bootstrapping still fails with:

regex.o: In function `extend_range_table_work_area':
/home/teirllm/emacscvsdir/emacs/lib-src/../src/regex.c:2069: undefined
reference to `xrealloc'
/home/teirllm/emacscvsdir/emacs/lib-src/../src/regex.c:2072: undefined
reference to `xmalloc'
regex.o: In function `regcomp':
/home/teirllm/emacscvsdir/emacs/lib-src/../src/regex.c:6295: undefined
reference to `xmalloc'
/home/teirllm/emacscvsdir/emacs/lib-src/../src/regex.c:6301: undefined
reference to `xmalloc'
collect2: ld returned 1 exit status
make[3]: *** [etags] Error 1
make[3]: Leaving directory `/home/teirllm/emacscvsdir/emacs/lib-src'
make[2]: *** [lib-src] Error 2
make[2]: Leaving directory `/home/teirllm/emacscvsdir/emacs'
make[1]: *** [bootstrap-build] Error 2
make[1]: Leaving directory `/home/teirllm/emacscvsdir/emacs'
make: *** [bootstrap] Error 2

lisp.h is, at least conditionally, already included in regex.c

Sincerely,

Luc.

Re: Recent changes to regex.c break bootstrapping

[Jan D.]

Luc Teirlinck wrote:
> After bootstrapping failed due to an undefined reference to xalloc, I
> made the obvious correction to xmalloc and installed in CVS, but after
> that bootstrapping still fails with:
> 
> regex.o: In function `extend_range_table_work_area':
> /home/teirllm/emacscvsdir/emacs/lib-src/../src/regex.c:2069: undefined
> reference to `xrealloc'
> /home/teirllm/emacscvsdir/emacs/lib-src/../src/regex.c:2072: undefined
> reference to `xmalloc'
> regex.o: In function `regcomp':
> /home/teirllm/emacscvsdir/emacs/lib-src/../src/regex.c:6295: undefined
> reference to `xmalloc'
> /home/teirllm/emacscvsdir/emacs/lib-src/../src/regex.c:6301: undefined
> reference to `xmalloc'
> collect2: ld returned 1 exit status
> make[3]: *** [etags] Error 1
> make[3]: Leaving directory `/home/teirllm/emacscvsdir/emacs/lib-src'
> make[2]: *** [lib-src] Error 2
> make[2]: Leaving directory `/home/teirllm/emacscvsdir/emacs'
> make[1]: *** [bootstrap-build] Error 2
> make[1]: Leaving directory `/home/teirllm/emacscvsdir/emacs'
> make: *** [bootstrap] Error 2
> 
> lisp.h is, at least conditionally, already included in regex.c

No, it is only included if emacs is #define:d (-Demacs), and it is not when 
compiling from lib-src.  But what are we trying to do, use xrealloc or define 
xrealloc as realloc?

In the first case alloc.c needs to be linked in also, or perhaps 
xrealloc/xmalloc defined in regex.c also (as it is in termcap.s for example).

	Jan D.

Re: Recent changes to regex.c break bootstrapping

[Andreas Schwab]

"Jan D." <jan.h.d@swipnet.se> writes:

> No, it is only included if emacs is #define:d (-Demacs), and it is not when 
> compiling from lib-src.  But what are we trying to do, use xrealloc or define 
> xrealloc as realloc?

IMHO this should just be reverted.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux Products GmbH, Maxfeldstraße 5, 90409 Nürnberg, Germany
PGP key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: Recent changes to regex.c break bootstrapping

[Chong Yidong]

> No, it is only included if emacs is #define:d (-Demacs), and it is not
> when compiling from lib-src.  But what are we trying to do, use
> xrealloc or define xrealloc as realloc?
>
> In the first case alloc.c needs to be linked in also, or perhaps
> xrealloc/xmalloc defined in regex.c also (as it is in termcap.s for
> example).

I went ahead and added xmalloc/xrealloc definitions for the non-emacs
part of regex.c.

Re: Recent changes to regex.c break bootstrapping

[Chong Yidong]

The 2006-02-19 change to regex.c, changing malloc and realloc to
xmalloc and xrealloc, doesn't do what it is apparently supposed to do,
since regex.c defines malloc as xmalloc, realloc as xrealloc, and free
as xfree when linked with Emacs.  For the case when it is not linked
with Emacs, I added xmalloc and xrealloc definitions (no need for
xfree, since its only purpose is to call BLOCK_INPUT), and put in the
same malloc -> xmalloc, realloc -> xrealloc redefinitions.

I hope this is correct way to go.  One worry: if another
non-Emacs-linked file #includes regex.c and defines its own xmalloc,
ugliness may ensue.

Re: Recent changes to regex.c break bootstrapping

[Giorgos Keramidas]

On 2006-02-20 12:08, Chong Yidong <cyd@stupidchicken.com> wrote:
> The 2006-02-19 change to regex.c, changing malloc and realloc to
> xmalloc and xrealloc, doesn't do what it is apparently supposed to do,
> since regex.c defines malloc as xmalloc, realloc as xrealloc, and free
> as xfree when linked with Emacs.  For the case when it is not linked
> with Emacs, I added xmalloc and xrealloc definitions (no need for
> xfree, since its only purpose is to call BLOCK_INPUT), and put in the
> same malloc -> xmalloc, realloc -> xrealloc redefinitions.
>
> I hope this is correct way to go.  One worry: if another
> non-Emacs-linked file #includes regex.c and defines its own xmalloc,
> ugliness may ensue.

If the intent is to make regex.[ch] as self-reliant as possible,
it may be better to copy xrealloc/xmalloc into regex.c and make
them static.

The etags build is easy to fix with a small patch to etags.c, which
I posted earlier, but I'm not sure if it's ok for regex.c to rely on
its callers to support xrealloc() and xmalloc().

Re: Recent changes to regex.c break bootstrapping

[Richard M. Stallman]

My change to regex.c turns out to be incorrect, and unnecessary,
since malloc is redefined as xmalloc when compiling for Emacs.