Log for #openttdcoop.devzone on 19th July 2010:
Times are UTC Toggle Colours
01:28:49  *** PeterT has quit IRC
01:29:32  *** PeterT has joined #openttdcoop.devzone
03:15:53  *** Frankr has quit IRC
08:16:42  *** Alberth has joined #openttdcoop.devzone
10:16:50  *** KenjiE20 has joined #openttdcoop.devzone
12:19:08  *** Seberoth has joined #openttdcoop.devzone
15:49:42  *** pm-quickview has joined #openttdcoop.devzone
15:56:35  *** Frankr has joined #openttdcoop.devzone
15:57:46  *** pm-quickview has quit IRC
16:13:26  *** frosch123 has joined #openttdcoop.devzone
16:21:06  <Brot6> grfcodec: update from r173 to r175 done -
16:21:44  *** welshdragon has quit IRC
16:23:59  <Brot6> nforenum: update from r381 to r384 done -
16:25:02  <Rubidium> no error logs... good!
16:25:33  <Brot6> nml: compile of r560 failed -
16:25:43  <Brot6> Following repos didn't need a nightlies update: 2cctrainset (r573), 32bpp-extra (r36), airportsplus (r52), bros (r12), comic-houses (r70), firs (r1074), fish (r386), heqs (r355), newgrf_makefile (r124), nmts (r16), nutracks (r86), ogfxplus (r40), opengfx (r467), openmsx (r90), opensfx (r96), snowlinemod (r15), swedishrails (r140), worldairlinersset (r655)
16:29:04  <Brot6> opengfx: rebuild of r467 done -
16:46:33  *** welshdragon has joined #openttdcoop.devzone
16:49:05  <Ammler> why did it recompile nml :-(
16:55:10  *** welshdragon has quit IRC
16:59:58  <Rubidium> because r560 has never been compiled yet?
17:00:06  <Rubidium> r559 is the last
17:03:00  <Ammler> yes, but r560 failed already yesterday
17:03:18  <Ammler> I guess, my rebuild update does somehow now skip the error rev check
17:03:42  <Rubidium> just leave it as is :)
17:04:20  <Alberth> url is also broken
17:04:29  <Ammler> hmm, since we use trunk, we could use REST to create a tickets on failed compile
17:26:04  <Brot6> NewGRF Meta Language - Revision 561:f8e87eb6379e: Fix (r560): Add missing colon. (Alberth) @
17:34:44  *** welshdragon has joined #openttdcoop.devzone
17:37:32  <Alberth> one daily failing attempt would do nicely
17:39:01  <Rubidium> but... it failed twice :)
17:45:00  <Ammler> Alberth: you fixed too fast
17:45:13  <Ammler> now I need another project to test the scheduler :-)
17:45:49  <Ammler> oh well, I postpone it...
17:46:27  <Alberth> next time I'll ask your permission to fix a compile problem :)
17:46:40  <Ammler> :-P
17:47:05  <Ammler> I am sure, this will happen soon
17:47:18  <Ammler> as you forget about regression tests quite fast
17:47:49  <Alberth> I run them as a standard procedure before commit/push
17:48:17  <Ammler> how do you use "you" unpersonalized?
17:48:45  <Alberth> 'one' or 'someone' ?
17:48:59  <Ammler> but I mean you :-)
17:49:00  <Rubidium> one can't help you with that
17:49:02  <Alberth> it is just a lot of typing:    make -C regression
17:49:44  <Ammler> "one of you" maybe
17:49:55  <Alberth> one of the developers :)
17:50:07  <Ammler> lazy dev
17:50:11  <Rubidium> if in doubt, rewrite :)
17:50:44  <Ammler> but it is good to see, that the regression tests helps
17:53:52  <Alberth> don't overestimate the usefulness. 'syntax error' like here is caught by almost any test. For more subtle things, you need a much larger test set, but that also takes more time to run.
17:54:45  <Ammler> well, you can make such test for the server
17:54:59  <Ammler> the server wouldn't care about time
17:55:45  <Ammler> make quick -C regression
18:03:13  <Brot6> GRFCodec - Feature #1108 (New): show revision number instead (or beside) hash (Ammler) @
18:03:49  <Rubidium> someone is REALLY anxious
18:04:21  <Ammler> just so you don't forget... :-P
18:04:37  <Ammler> (you=you)
18:07:15  <Rubidium> not that repo managers are encouraged to package nightlies
18:08:02  <Ammler> yes, but when you run <tool> --version, you don't see, which nightly you have
18:08:09  <Ammler> you need to use your package manager
18:09:00  <Ammler> oh, are you planning to make regular releases of nforenum and grfcodec, when it gets official?
18:10:05  <Ammler> I don't think, there is a package manager using release grfcodec
18:10:41  <Rubidium> no, there isn't one yet... because the last release of grfcodec is totally and utterly useless
18:10:53  <Ammler> or a nforenum, which autoupdates the dat files?
18:11:17  <Rubidium> like 4 years old and nforenum is also already more than 2.5 years old
18:11:39  <Rubidium> and the dat files only help you so far...
18:11:49  <Rubidium> adding new features are stuff and it kinda breaks horribly
18:12:21  <Ammler> well, maybe it would be better to trash the "dat cache" then?
18:12:39  <Rubidium> I'm not quite sure about the workings of that "feature"
18:12:47  <Rubidium> so... waiting for Yexo on that
18:13:13  <Ammler> it is not downwards compatible, as we already knew :-)
18:15:44  <Rubidium> releasing a grfcodec / nforenum every 6 or so months should be ample
18:16:16  <Rubidium> somewhere around christmas to prepare for the new OpenTTD release and somewhere in summer so there are reasonably frequent updates
18:16:26  <Rubidium> or when there's a serious issue that needs addressing
18:16:45  <Rubidium> and if people need a newer "feature", they should "just" use the nightlies that will still be compiled
18:19:26  <Ammler> yes, but specially on nightlies, it would be nice to see, which is newer
18:20:13  <Ammler> timestamps might be hard to keep
18:21:34  <Rubidium> YES
18:21:43  <Rubidium> stop whining on about it
18:22:33  <Ammler> :-)
18:23:27  <Ammler> hmm, do I also need to rebuild releases, when dependencies change?
18:23:53  <Rubidium> I wouldn't do that
18:24:31  <Rubidium> and actually rebuilding and giving it the same name might not be that wise either; people might think they've got a compromised binary if the md5 checksum doesn't match anymore
18:25:20  <Ammler> he, I wait for that moment
18:26:02  <Ammler> It didn't happen so far, that md5 changed during other version
18:26:33  <Rubidium> I'm talking about the md5 checksum of the downloadable packages
18:26:46  <Ammler> maybe that should be marked as failed and keep the last release
18:26:58  <Ammler> oh
18:27:07  <Ammler> nobody does compare those :-)
18:27:41  <Ammler> I wonder, if there are some automatic scripts doing it...
18:28:15  <Ammler> you compare the sums, when something fails)
18:29:14  <frosch123> package managers do that
18:30:41  <frosch123> emerging java docs on gentoo gives you a hint, that if the checksums fail, sun likely uploaded a new version without changing the filename, and that you shall notify gentoo in that case to update the checksums :)
18:31:35  <frosch123> but for most stuff the distributor has its own mirrors, so..
18:31:57  <Ammler> don't they check gpg keys?
18:32:25  <Ammler> or are those for the repos only
18:32:25  <frosch123> for what?
18:32:27  <Rubidium> true... but claiming you uploaded the official source tarball and then the checksums don't match the official tarball is a sure way to get into a huge shit storm
18:33:29  <Ammler> Rubidium: then we should rebuild and if the diff is empty keep the old and trash the rebuild
18:33:53  <Ammler> and add RELEASE number to the filename
18:34:35  <Ammler> but well, as said, nobody does use the nightly for official packages
18:36:53  <Ammler> around 90% of my system updates are simple rebuilds
18:44:15  <Alberth> and if the diff is not empty, keep the old and trash the rebuild?
18:49:36  <Ammler> if the diff isn't empty, fail the build
18:50:29  <Ammler> basically buidling for comparison only
18:50:36  <Ammler> never keep the rebuild
18:52:40  <Alberth> I don't really understand how anything released can change anymore, but if you say so, ok
18:53:21  <Ammler> well, it changes the timestamps for example
18:53:50  <Ammler> that allways, but if you optimize something on nml, it can also change the grf
18:54:37  <Alberth> a changed timestamp already breaks the md5sum, sin ce the tar file changes
18:54:39  <Ammler> that didn't happen since we have rebuild support
18:54:50  <Ammler> Alberth: exactly
18:56:04  <Alberth> I'd consider rebuilding a release dangerous, you never know what new problems you introduce.
18:56:26  <Ammler> Alberth: suse does that all the time
18:56:39  <Ammler> I am sure, other distros do that too
18:57:15  <Alberth> I don't see how that makes it ok tbh.
18:57:46  <Ammler> but we might need to add the RELEASE number or keep the old, that is what I wonder.
18:58:21  <Alberth> I'd say build once, never touch it again.
18:58:47  <Ammler> rebuild is fine, just what doing with it is the question :-)
19:01:36  <Alberth> I see no use at all.  A release is a snap-shot at some point in time, just like a nightly or a revision.
19:02:01  <Alberth> but I am not so much in favor of using eg rebase :)
19:02:26  <Ammler> well, it would matter, if someone builds a release with nml now and the resulting md5sum does differ from the one on bananas
19:02:55  <Alberth> not if he uses the exact same nml revision
19:03:06  <Ammler> how does "he" know?
19:04:05  <Alberth> probably not, just like you have no clue what c library and c compiler you use, and what python to run nml
19:04:41  <Ammler> well, with rebuild, you would get notice about it
19:05:28  <Alberth> and then?
19:05:36  <Ammler> dunno :-)
19:05:39  <Alberth> you ask yexo to revert the change in nml?
19:05:44  <Ammler> maybe, new release
19:06:46  <Alberth> at some point, your builds will even fail
19:07:00  <Alberth> eg try compiling openttd a few thousand revisions ago
19:07:16  <Alberth> todays compilers don't eat that code any more.
19:07:42  <Ammler> well, you only rebuild last release
20:00:38  *** Alberth has left #openttdcoop.devzone
20:02:20  *** PeterT has quit IRC
20:02:54  *** PeterT has joined #openttdcoop.devzone
20:51:26  <Brot6> HEQS "Heavy Equipment" Set - Revision 356:d1d3022cf486: Change: tweaks to code for Gmund Mog (andythenorth) @
20:54:19  *** frosch123 has quit IRC
20:56:31  <Brot6> World Airliners Set - Revision 656:b38600d1adf4: 7F expanded and Airbus changes (Frank) @
21:47:07  *** Seberoth has quit IRC

Powered by YARRSTE version: svn-trunk