-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Jérôme Marant <[EMAIL PROTECTED]> writes:
> Roger Leigh <[EMAIL PROTECTED]> writes: > >>> Timing-dependent? How so? >> >> Running another build concurrently, or using strace (-f -F), or 'make >> -d' would make an otherwise failing build succeed. I'm fairly >> confident that timestamp skew is causing at least part of the problem. >> Those two changes mitigate this, somehow. Without being slowed down >> in this way, temacs gets rebuilt and this then causes fns-21.4.2.el to >> be created in addition to fns-21.4.1.el, which then causes the build >> to fail as mentioned in the bug report. >> >> Check for config.status being run in: >> http://people.debian.org/~rleigh/emacs21_21.4a-1.build-log [succeeds] >> http://people.debian.org/~rleigh/emacs21_21.4a-1.build-nochroot-log [fails] > > I'm sorry, but I don't quite get it. I guess I'm tired. By reading logs, > I can't identify the place causing the trouble. Are we talking about > parrallel builds? Otherwise I can't understand how running things > sequentially would cause race conditions. Running /any/ other build or cpu/disk hogging process on the system slows down the build just enough to prevent the race condition occuring. In the above logs, search for "configure: creating ./config.status". You'll see there are more instances in the failing build. Some unknown cause caused configure to be re-run. This cause (currently unknown) is, as far as we can tell, some race causing incorrect (skewed) timestamps, which confuse make into thinking some target is not up-to-date. [I tell a lie: the cause is apparent from reading the root Makefile.in, but what triggers that is not obvious.] The real challenge is finding what the initial trigger is. Since debugging alters the timing, the race vanishes, and the build succeeds... Regards, Roger - -- Roger Leigh Printing on GNU/Linux? http://gimp-print.sourceforge.net/ Debian GNU/Linux http://www.debian.org/ GPG Public Key: 0x25BFB848. Please sign and encrypt your mail. -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.1 (GNU/Linux) Comment: Processed by Mailcrypt 3.5.8 <http://mailcrypt.sourceforge.net/> iD8DBQFCpLtWVcFcaSW/uEgRAgGeAJ9KrrZWlQhGYiuaqgIKxWrzFebRmACg1u2Y ETdJkVTT+u3ScfjdNRNO2yc= =O/// -----END PGP SIGNATURE-----
