Opened at 2010-11-12T14:10:23Z
Closed at 2011-08-26T22:20:20Z
#1257 closed defect (invalid)
the "update-apt" step for pycryptopp buildbots always says that a task is already queued
Reported by: | zooko | Owned by: | somebody |
---|---|---|---|
Priority: | major | Milestone: | undecided |
Component: | dev-infrastructure | Version: | 1.8.0 |
Keywords: | debian packaging buildbot | Cc: | warner |
Launchpad Bug: |
Description
Here is a builder whose job is to update the apt repository:
http://tahoe-lafs.org/buildbot-pycryptopp/builders/Arthur%20debian-lenny-c7-i386%20syslib/
It seems like every time the update-apt step is run, it says:
update already queued, deferring to them
stderr_dk on IRC says:
<stderr_dk> I wonder where http://tahoe-lafs.org/buildbot-pycryptopp/builders/Arthur%20debian-lenny-c7-i386%20syslib/builds/7 were uploaded... <stderr_dk> It says "python-pycryptopp_0.5.25-r755-1_i386.deb: uploaded" <stderr_dk> But the latest version in http://tahoe-lafs.org/debian/dists/lenny/main/binary-i386/ is 0.5.17 from 01-Jun-2010. <stderr_dk> Maybe the .deb is uploaded to the wrong place? Maybe that's part of the problem?
I think the reason it isn't served up by http://tahoe-lafs.org is that the script which always thinks it is already queued is supposed to scp it to tahoe-lafs.org.
Change History (7)
comment:1 Changed at 2010-11-12T14:17:11Z by zooko
comment:2 Changed at 2010-11-12T14:22:11Z by zooko
Reading the Makefile I don't see any flaw in its handling of queue.stamp. I guess what happened was that the system rebooted or in some other way the make process exited without finishing its clean-up of those .stamp files. I'm going to manually remove them.
comment:3 Changed at 2010-11-12T14:34:59Z by zooko
That fixed it! But there is a new failure: http://tahoe-lafs.org/buildbot-pycryptopp/builders/Arthur%20debian-lenny-c7-i386%20syslib/builds/10/steps/update%20apt%20repo/logs/stdio
make[3]: Leaving directory `/home/buildslave/tahoe-debs' ~/bin/mirror-debs-to-org.sh pushing .debs to tahoe-lafs.org Host key verification failed. make[2]: Leaving directory `/home/buildslave/tahoe-debs' make[1]: Leaving directory `/home/buildslave/tahoe-debs' rsync: connection unexpectedly closed (0 bytes received so far) [sender] rsync error: unexplained error (code 255) at io.c(601) [sender=3.0.7] make[2]: *** [run-update-and-mirror] Error 255 make[1]: *** [run-update-and-check-for-queued] Error 2 make: *** [update] Error 2 program finished with exit code 2
comment:4 Changed at 2010-11-12T14:49:37Z by zooko
I suspect it might have just been waiting for someone to type "yes" to ssh. I did so manually. I guess this means we should be configuring our ssh to not prompt that. There is a way to do that.
comment:5 Changed at 2010-11-12T15:14:29Z by zooko
Aha, and that led to http://tahoe-lafs.org/buildbot-pycryptopp/builders/Arthur%20debian-lenny-c7-i386%20syslib/builds/11/steps/update%20apt%20repo/logs/stdio
update in progress, queueing a new one
So I guess that timeout from scp waiting for someone to type "yes" is what led to the leftover stale queue.stamp file...
comment:6 Changed at 2010-11-12T17:16:09Z by zooko
Okay that worked: http://tahoe-lafs.org/buildbot-pycryptopp/builders/Arthur%20debian-lenny-c7-i386%20syslib/builds/12/steps/update%20apt%20repo/logs/stdio
Now http://tahoe-lafs.org/debian/dists/lenny/main/binary-i386/?C=M;O=D has the latest files.
To fix this ticket: make it so failures of the upload code don't leave behind an immortal stale lock file (probably by using the debian/ubuntu package lockfile-progs and perhaps the bash trap builtin or else a subprocess).
Alternate fix for this ticket: decide that we can live with that script doing that whenever it dies and close this ticket as "fixed".
comment:7 Changed at 2011-08-26T22:20:20Z by davidsarah
- Resolution set to invalid
- Status changed from new to closed
The deb builders have been decommissioned.
I found the script on dev.allmydata.org and ran it manually:
inspecting the Makefile I see that it is looking for a file named queue.stamp: