Subject: pkg/26887: pkgsrc is too fast for print-PLIST
To: None <gnats-bugs@gnats.NetBSD.org>
From: None <firstname.lastname@example.org>
Date: 09/08/2004 15:26:34
>Synopsis: pkgsrc is too fast for print-PLIST
>Arrival-Date: Wed Sep 08 15:27:00 UTC 2004
>Originator: Gavan Fantom
>Release: NetBSD 2.0_BETA
NetBSD dhcp-ewat02-22 2.0_BETA NetBSD 2.0_BETA (HP) #3: Wed Aug 11 17:02:25 BST 2004 gavan@dhcp-ewat02-22:/usr/src/sys/arch/i386/compile/HP i386
When executed on a fast machine, the print-plist regression test intermittently fails because "make print-PLIST" fails to produce correct output.
This is apparently due to the fact that pkgsrc is so blindingly fast that the extract cookie and the test file created by the regression test package have the same timestamp. print-PLIST only reports files strictly *newer* than the extract cookie, and so fails in this case.
This problem will get worse as faster machines become available.
Buy a fast machine. Install pkg_regress and run it lots of times. Watch the print-plist test intermittently fail.
If it doesn't fail, repeat with faster machine.
The regression test can easily be made to stop reporting this problem by inserting a "sleep 1" in the do-build target.
Other workarounds include building KDE in the background and reducing the system clock speed in the BIOS.
The ideal fix would be to teach find how to check for files that have a greater or equal timestamp.