Re: pkgconfig-0.5.0

[Havoc P. <hp...@re...>]

Martijn van Beers <ma...@ee...> writes: 
> Ok, this is supposed to be caught by a check for flockfile in your
> libc. Seems like you have one that does have flockfile, but not
> getc_unlocked. or at least not without a #define.
> 
> Could you tell us exactly which version of libc and what distro
> you are running? thanks,
> 

<tangents>
Can we please put the ChangeLog back? Emacs will auto-format a
ChangeLog, it won't auto-format CVS stuff, and I don't want to wait on
a CVS command every time I want to grep ChangeLog. Also ChangeLog
should be in EXTRA_DIST so tarball recipients can see what changed.
cvs2cl.pl is not in CVS that I see, so I can't use it. And 'cvs log'
does not easily show changes for the whole directory, it just shows
changes file-by-file. Finally it is a standard convention to use
ChangeLog and doing something different is just gratuitous.

In Makefile.am, m4_SCRIPTS is wrong, it should be m4_DATA. I thought I
already changed this once? What is the problem you're trying to solve?
SCRIPTS is for things with the execute bit set. m4 macros are macros,
not scripts. There is no +x bit.

PKG_CHECK_CFLAG has no business in pkg.m4, neither does
PKG_ACLOCALFLAGS. PKG_CHECK_CFLAG should be in autoconf. Send them a
patch. PKG_ACLOCALFLAGS ditto. In the meantime a little cut-and-paste
won't hurt anyone, configure.in are full of that anyway. PKG_ should
be macros involving packages, not random-ass whatever we can think of.

PKG_CHECK_MODULES m4 macro is still checking for 0.7, but you've
shipped 0.6. Not going to work... The reason for 0.7 was to bump right
before release since some people were using CVS version. So this way
we catch potential issues. But if you want to go back to 0.6 for
aesthetics, you need to fix the m4 macro.

I had to dig through a diff to find all these changes, due to lack of
ChangeLog. Also, the log messages logged to CVS are in the wrong
format; which I'll add to the reasons ChangeLog should be used, people
always use lame CVS log messages in a lame format otherwise.

Finally, I need a release with this stuff not broken! Can we
re-release please? Or if you don't have time, give me the right access
to sourceforge and I'll do it.
</tangents>

Re: getc_unlocked(), why not try just sticking this #define at the top
of that file, release, see if problems still come up, if the issue
comes up again let's just remove flockfile() uses and
s/getc_unlocked/getc/g for the next release.

Havoc