Re: [AD] Some Debian compatibility issues

[Shawn Hargreaves <shawn@xxxxxxxxxx>]

Grzegorz Adam Hankiewicz <gradha@xxxxxxxxxx> writes:
> configure.in:127: warning: AC_TRY_RUN called without default to allow 
> cross compiling

That's not a problem, or at least if it is a problem, the bug is in the 
autoconf macros rather than in our side of things, since that's one of the 
standard macros that ships with autoconf (it's the bit that checks the 
processor endianess). It can safely be ignored. Would be nice to get rid of, 
but I have no idea how!

> Then, the un/install targets aren't compatible with debian's install-info.
> I have tried to figure out from where does the makefile build itself, but 
> I am lost...

The configure script creates it from makefile.in, changing all the 
@variable@ bits depending on what options it detected. To change things, 
edit makefile.in and then run configure again, or if you need to change the 
actions of configure itself, edit configure.in and then run autoconf to 
create a new configure script.

> Also, is there's some way configure could check if the install-info 
> program is from Debian, I can provide you with a working setup of the 
> needed switches if the Debian install-info is detected.

My "install-info --version" prints:

   install-info (GNU texinfo) 3.12f

   Copyright (C) 1999 Free Software Foundation, Inc.
   There is NO warranty.  You may redistribute this software
   under the terms of the GNU General Public License.
   For more information about these matters, see the files named COPYING.

What does yours have to say for itself? Also, does the makefile actually try 
to run install-info and gets an error, or does it skip that entirely? (it 
already has a check that tries running install-info --version and then gives 
up if that fails). If your version message is different somehow, perhaps it 
could grep the results of that and then use different options depending 
which it finds. That is somewhat ridiculous, though: it strikes me that an 
installation utility is one thing that absolutely has to stay compatible, 
and yet my version of it won't work at all with the options that yours 
accepts!

> Now I tried a compile with --enable-strict-warnings, and I got:

Yeah, I had all sorts of problems with that. Some crazy warnings about 
memcmp() as well, which couldn't be fixed no matter how I tried to cast the 
arguments. And the configure script itself goes haywire if you try to run it 
with -Wall -Werror, as half the test programs it uses will then fail, so it 
ends up deciding that half of libc is missing! It looks to me like either 
the libc headers or the autoconf test programs, or maybe both, just aren't 
robust enough to handle that level of strictitude...


--
Shawn Hargreaves - shawn@xxxxxxxxxx - http://www.talula.demon.co.uk/
"A binary is barely software: it's more like hardware on a floppy disk."