summaryrefslogtreecommitdiff
AgeCommit message (Collapse)AuthorFilesLines
2012-03-26Fix date in NEWS fileobnam-0.26Lars Wirzenius1-1/+1
2012-03-26debian/control: Bump Standards-Version to 3.9.3. No other changes needed.Lars Wirzenius2-1/+2
2012-03-26Only run fast checks during Debian package buildLars Wirzenius2-2/+4
2012-03-25Only check metadata for changesLars Wirzenius1-2/+2
The other files may change, because of larch journalling code.
2012-03-25Release version 0.26Lars Wirzenius6-6/+14
2012-03-25Fix repository corruption problems due to crashingLars Wirzenius5-1/+109
2012-03-25Fix exception creation by adding missing messageLars Wirzenius1-1/+1
2012-03-24Oops, had forgotten to add the crash-test scriptLars Wirzenius1-0/+73
2012-03-24Add crash-test to setup.pyLars Wirzenius1-0/+3
2012-03-23Add more crash pointsLars Wirzenius1-1/+7
2012-03-17Fix setting nameLars Wirzenius1-1/+1
2012-03-17Add --crash-limit optionLars Wirzenius1-0/+10
2012-03-17Add optional crashing to LocalFSLars Wirzenius1-0/+16
This makes perfect sense, for crash testing.
2012-03-14Remove --wiki option from obnam-benchmarkLars Wirzenius1-4/+0
2012-03-11Add locking tests to setup.py checkLars Wirzenius1-0/+16
2012-03-10Update manual page about lockingLars Wirzenius1-0/+19
2012-03-06Add repository locking between clientsLars Wirzenius19-87/+570
2012-03-06Allow user to specify number of generations to testLars Wirzenius1-5/+10
2012-03-06Update NEWS about new featureLars Wirzenius1-0/+7
2012-03-06Re-open repository after committing checkpointLars Wirzenius1-0/+1
This makes sure there are no lingering stale cached B-tree nodes in memory anymore.
2012-03-05Allow access to backup repository via sftpLars Wirzenius3-20/+22
Separate the URL for obnam's access to the repository, and the directory pathname for the test scripts' access.
2012-03-05Refactor to hide magic number behind variable with descriptive nameLars Wirzenius1-1/+3
2012-03-04Increate number of test generations for locking to 16Lars Wirzenius1-1/+1
This didn't actually find any bugs. I'm disappoint.
2012-03-03Fix problem with adding clients during backupLars Wirzenius1-5/+10
Found a problem where client-1 adds itself to the repository, and then client-2 does the same, but client-2 had already read the B-tree into memory before client-1 added itself, and so when client-2 saved the modified B-tree, it accidentally overwrote the changes client-1 had made, resulting in no client-1 in the repository. Oops.
2012-03-03Be a bit more verbose in the verification testLars Wirzenius0-0/+0
This is all about making it easier to debug problems. Which I'm having now. In case you didn't guess. Nobody ever reads my commit messages, so you don't even exist. :)
2012-03-03Add tracing to allow easier debugging of problemsLars Wirzenius1-0/+1
2012-03-03Add tracing to log files, so it'll be easier to debug problemsLars Wirzenius0-0/+0
2012-03-03Run tests with 8 generations instad of 2Lars Wirzenius1-1/+1
This will uncover more exciting bugs!
2012-03-03Handle obnam read-only errors sanelyLars Wirzenius1-5/+16
Read-only errors need to be handled by re-trying. Now we do that.
2012-03-03Do not run test-many-generations with set -e, so we can catch errorsLars Wirzenius1-1/+1
2012-03-03During lock testing, handle obnam restore failures gracefullyLars Wirzenius2-2/+22
Because of how Obnam deals with locking (it only locks for writes), read-only operations such as restore can fail because a backup is happening concurrently. We deal with this by re-trying the restore until it succeeds (i.e., obnam does not crash). There's a chance of infinite looping if the crash happens for another reason, but that's for the human watching the test to deal with.
2012-03-03Fix creation of LocalFS baseurl race conditionLars Wirzenius1-1/+7
If we have many clients running concurrently, they might all be creating, say, the chunks directory, and with the old code only one of them would succeed, and the others would fail, for not good reason. The new could should deal with that.
2012-03-02Save output from childrenLars Wirzenius1-1/+1
This aids debugging something wonderful.
2012-03-01Test two generations per client, instead of just oneLars Wirzenius1-1/+1
To catch more bugs.
2012-03-01Fix write_file for LocalFSLars Wirzenius1-8/+2
The previous code had, I think, been moved from generic VFS code, where it had to deal with SFTP limitations. Since the code is now LocalFS specific, removing the documented race condition can only be regarded as overdue. (Meanwhile, the actual SFTP code seems to work better.)
2012-03-01Lock first, check format version afterwards.Lars Wirzenius1-1/+1
2012-03-01Store golden summain files with per-client namesLars Wirzenius1-3/+5
Previously, every client used the same names, which meant they overwrote each others' files. Oops.
2012-03-01Generate different and random contents every timeLars Wirzenius1-3/+3
Previously, the same data was being generated all the time, for every generation in every client. This meant that fewer bugs were caught. Also, reduce the output volume so that errors are more easily seen.
2012-02-27Add test tools for testing individual lock filesLars Wirzenius2-0/+100
2012-02-25Fix argument passing to the utimensat wrapperLars Wirzenius1-4/+10
A struct timespec does not have two long fields; the seconds are a time_t. We need to deal with that.
2012-02-25Fix LockManager._sleep to not be given argumentsLars Wirzenius2-4/+3
2012-02-25Implement --lock-timeoutLars Wirzenius3-12/+15
2012-02-25Set default for --lock-timeoutLars Wirzenius2-3/+5
2012-02-25Add --lock-timeout optionLars Wirzenius1-0/+6
2012-02-25Undo debugging thingLars Wirzenius1-1/+1
2012-02-25Add commit_shared and related changes to how locking happensLars Wirzenius7-9/+112
Fixes everywhere.
2012-02-25Add methods to lock/unlock shared directoriesLars Wirzenius2-0/+43
2012-02-25Use LockManager to create root, per-client locksLars Wirzenius1-18/+11
2012-02-25Sort directories to be lockedLars Wirzenius2-2/+9
No test for this, since it's locale-dependent and meh.
2012-02-25If one lock fails, do not lock anythingLars Wirzenius2-1/+15
Or rather, unlock the ones we already locked.
generated by cgit v1.2.1 (git 2.18.0) at 2024-05-10 20:02:35 +0000