Re: Re:

[Gary V . Vaughan]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1


On 1 Apr 2004, at 00:06, Bob Friesenhahn wrote:
> On Wed, 31 Mar 2004, Gary V.Vaughan wrote:
> > > The style of this change confuses me, and it doesn't seem to do
> > > exactly the same thing:
> > >
> > >    /* If ARGZ_LEN has shrunk to nothing, release ARGZ's memory.  */
> > >    if (!argz_len)
> > > -    LT_DLFREE (argz);
> > > +    argz = (free (argz), (char *) 0);
> >
> > Actually, they are exactly the same after cpp is done!
>
> Yes, it guess it is valid C code.

Okay, you got me... that code is gross.  It was gross in a macro 
before, but
now it is gross in the body of the code.  I guess we need an xfree to 
hide it.

> > Gnu programmers are used to the xmalloc api, and it is reimplemented 
> > all
> > over the place, and lowers the barrier for contributing, so I'm 
> > inclined
> > to reuse the idiom if possible.  When it comes to splitting out xalloc
>
> GNU programmers are not the only users of libtool and libltdl.

I meant to say that GNU programs that look like each other make it 
easier
for potential contributers to read them all.

> > functions, we can leave them static and #include "xalloc.c" into 
> > ltdl.c.
> > Eventually I'd like there to be enough sanity in libltdl source that
>
> It is really poor form to include .c files into .c files.  While it is
> true that a number of projects do that (FreeType and libwmf come to
> mind), it can cause problems for some tools, particularly automated
> Makefile generators or other tools that place significance on files
> based on their extension.

Yeah, agreed.

: dons flameproof armour

Now that we don't RTLD_GLOBAL anymore there is no reason for internal 
apis
to leak into libltdl clients... :-)

> > someone with a reasonable familiarity with the GNU programming style 
> > and
> > idioms could get to grips with ltdl.c in a few hours... the first step
> > is eliminating as much macro noise, and not-invented-here-isms as I 
> > can.
>
> In some ways the macroized version looks cleaner.

Okay, I went overboard.  I think HEAD has too much code between 
LT_EREALLOC
and sbrk.  I think my patch 100 has too little. Lets see if 100bis is 
less
controversial ;-)

Cheers,
        Gary.
- --
Gary V. Vaughan      ())_.  address@hidden,gnu.org}
Research Scientist   ( '/   http://tkd.kicks-ass.net
GNU Hacker           / )=   http://www.gnu.org/software/libtool
Technical Author   `(_~)_   http://sources.redhat.com/autobook

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.2 (Darwin)

iD8DBQFAa1UgFRMICSmD1gYRAvw/AJ9AB/WKdaWxBr1GwWreJGv1KmP5cACeNYpY
y3mgQN5bX3lDnF2Kn0vSEKY=
=xgPA
-----END PGP SIGNATURE-----