Re: [bug-gawk] Question about git build system

[Assaf Gordon]

Thanks!

Aharon Robbins wrote, On 12/15/2012 03:18 PM:
> See the node "Derived Files" in doc/gawk.texi in the master branch for
> an explanation of why the derived files are there.
> 

That section indeed clears things up.

While I understand the need to keep the derived yacc/bison files in the 
released tarballs,
keeping the autotool derived files in the git is quite different from almost 
all common project.

The goal is to make like easier for the users and developers, but isn't the 
assumption that if the user cloned a git repository (as opposed to downloaded a 
tarball) - he/she is already familiar (or expected to be familiar) with 
autotools? and have everything in-place?

But as the comment in the manual says, this is the maintainer's reasoning and 
philosophy - so there's no point arguing :)

Regarding Eli's comment:

Eli Zaretskii wrote, On 12/15/2012 03:26 AM:
> 
> It's on purpose.  Does it give you trouble in any way?
> 
>> Compared with other autotools-based projects, it's a bit unexpected that 
>> running "./bootstrap.sh" does not regenerate "configure" .
> 
> Unexpected, maybe.  But does it do any harm?

It had two side-effects (which of course aren't too bad, just surprising):

1. When I changed "configure.ac", running "bootstrap.sh" didn't do "anything".
Perhaps it would help to provide an additional script to do what "bootstrap" 
was supposed to do? and include it only in the git repository, not in the 
released tarball?
===
$ cat reconf.sh 
#!/bin/sh

rm -f autom4te.cache aclocal.m4
aclocal -I m4
autoconf
autoheader
automake -a

cd extension
rm -f autom4te.cache aclocal.m4
aclocal -I m4
autoconf
autoheader
automake -a
cd ..
===

2. After I was able to re-generate "configure" and "Makefile.in" and all the 
other files, a tiny change in "configure.ac" resulted in a very large patch 
(basically, every autotool related file was modified) - so publishing that 
patch would be somewhat inefficient and confusing.


Regards,
 -gordon