[torquedev] missing file checks

Glen Beane glen.beane at gmail.com
Fri Jul 23 12:04:40 MDT 2010


On Fri, Jul 23, 2010 at 1:41 PM, Ken Nielson
<knielson at adaptivecomputing.com> wrote:
> On 07/23/2010 11:16 AM, Michael Jennings wrote:
>> On Friday, 23 July 2010, at 17:48:57 (+1000),
>> Chris Samuel wrote:
>>
>>
>>> My main worry is that without nightly snapshots with those generated
>>> files in them it becomes even harder to get people to test recent
>>> changes as to test they'll go have to go through the pain of trying
>>> to get a current version of autoconf, automake, etc working.
>>>
>>> I agree that they ideally they shouldn't be there, but we need to
>>> have something regularly produced that does have them for ease of
>>> testing.
>>>
>> I haven't checked, but it would be my hope that all snapshots would be
>> produced the same way a release would (should?) be produced:
>> ./configure&&  make distcheck
>>
>> Many projects have an "autogen.sh" script which does some variation of
>> the following:
>>
>> libtoolize -c -f&&  aclocal&&  autoconf&&  autoheader&&  automake -a -c&&  ./configure "$@"
>>
>> Or, for more modern autotools collections:
>>
>> autoreconf -i -s&&  ./configure "$@"
>>
>> In those cases, the snapshot command simply becomes ./autogen.sh&&
>> make distcheck.  Even without the script, autoreconf is still only one
>> additional command, and it only has to be run by the person or script
>> actually generating the snapshot, not by the user downloading the
>> snapshot.
>>
>> No tarball, snapshot or otherwise, should ever be produced without
>> containing the configure and Makefile.in files, libtool and
>> ltmain.sh scripts, aclocal.m4, config.h.in, etc.  We're strictly
>> talking about what the repository itself contains.
>>
>> Michael
>>
>>
> So what does everyone do currently with the TORQUE tarball. When you
> install and configure TORQUE do you run autoconf and automake in a
> autogen.sh manner and then run configure. Or do you simply take what's
> there and run ./configure?
>
> I am for whatever makes life easier for everyone.

I'm not sure why anyone that downloads a .tgz would want to rerun the
autotools, unless they wanted to add some source files or something
like that

If we no longer check in the generated files the only people that
would run the autotools would be people that grab source straight from
subversion.  There is no need for someone downloading a snapshot or
release to do anything other than ./configure - the generated files
get included in distributions and snapshots


More information about the torquedev mailing list