[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: experimental tex-common packages with triggers support



Michael Banck <mbanck@debian.org> wrote:

> Hi,
>
> On Wed, Jul 02, 2008 at 05:19:47PM +0200, Hilmar Preusse wrote:
>> On 01.07.08 Michael Banck (mbanck@debian.org) wrote:
>> > Norbert Preining wrote:
>> > > I have uploaded tex-common 1.12 to experimental. This release
>> > > contains triggers support. But as long as packages shipping
>> > > updmap.d and language.d files are rebuild we wont profit.
>> > 
>> > That's great, but I'm wondering whether this also includes trigger
>> > support for mktexlsr?
>> > 
>> IIRC Norbert decided not to implement triggers for mktexlsr for now,
>> as the command does not need that much system resources.
>
> Hrm, my mail was mainly motivated by the gprolog build on the Debian
> GNU/Hurd buildd.  It installed half/all of texlive, failed very early in
> configure, and removed all of it.  It took more then 24 *hours* to do
> that, and as I saw those "mktexlsr" messages for most of that time, I
> assumed that was the culprit, maybe.  It could be that mktexslr does
> something which turns out to be pathological on hurd-i386 only.

Didn't it call updmap, too?  That takes much more time than
mktexlsr. Format generation is also rather slow.

> Well, the usual rules for experimental apply - few people use it unless
> really pushed for it.  If you did not get any feedback, assuming that
> everything is fine is not warranted.  On the other hand, as I guess you
> can easily back out the feature again in case of trouble, uploading it
> to unstable now that you did not get any negative feedback (and possibly
> revert it in case you then get some) looks alright to me.

Backing out is not so easy. If we upload to unstable, but people don't
upload packages to unstable which use the feature, it's just as untested
as in experimental. If they do, we cannot simply drop it, unless they
kept their manual calls in the postinst - which means that a) the time
needed would be doubled, and b) it's not tested whether the trigger run
is in fact sufficient, run at the right time etc.

Regards, Frank

-- 
Frank Küster
Debian Developer (teTeX/TeXLive)


Reply to: