559
|
|
|
Michał Sawicz |
9 years ago
|
|
|
558
|
|
|
Michał Sawicz |
9 years ago
|
|
|
557
|
|
|
Pawel Stolowski |
9 years ago
|
|
|
556
|
|
|
Pawel Stolowski |
9 years ago
|
|
|
555
|
|
|
Marcus Tomlinson |
9 years ago
|
|
|
554
|
|
|
Michi Henning |
9 years ago
|
|
|
553
|
|
|
Michi Henning |
9 years ago
|
|
|
552
|
|
|
Marcus Tomlinson |
9 years ago
|
|
|
551
|
|
|
Marcus Tomlinson |
9 years ago
|
|
|
550
|
|
|
Pawel Stolowski |
9 years ago
|
|
|
549
|
|
|
Michi Henning |
9 years ago
|
|
|
548
|
|
|
Michi Henning |
9 years ago
|
|
|
547
|
|
|
Michi Henning |
9 years ago
|
|
|
546
|
|
|
CI bot |
9 years ago
|
|
|
545
|
|
|
Marcus Tomlinson |
9 years ago
|
|
|
544
|
|
|
Pawel Stolowski |
9 years ago
|
|
|
543
|
|
|
Michi Henning |
9 years ago
|
|
|
542
|
|
|
Michi Henning |
9 years ago
|
|
|
541
|
|
More detail in ScopesWatcher trace as to what's going on. Removed watches for .so and settings.ini files. They are not needed: if the .so is updated without the .ini file being updated, we'll pick it up the next time the scope is started. If the .so is removed but the .ini is left behind, the scope won't start because it is incorrectly installed, and we'll get a failure from the scopeloader. Settings files also don't need watching because the updgrade for a scope will re-write all files, not just the settings.ini. Changed inotify code to remove triggering on IN_ATTRIB, because we don't care about that. Added trigger on IN_CLOSE_WRITE, so we do get a notification when a .ini file is copied into the scope install dir and has finished updating. Changed code in ScopesWatcher to not emit a notification if a .ini file is created, but is empty. This avoids a race where we receive the CREATE event and try to parse the file contents before they have been written. A few other minor fixes: watch_event() is now pure virtual in DirWatcher, better error checking for data read from inotify fd, improved comments. Fixed tests to not wait for 1 second in the hope that this will be long enough and producing bogus failures on a slow machine. Instead, the test now busy-wait for up to five seconds. If things finish faster, the test finishes faster, and we wait for five seconds if things really aren't working. Fixes: https://bugs.launchpad.net/bugs/1394080.
Approved by Marcus Tomlinson, PS Jenkins bot.
|
Michi Henning |
9 years ago
|
|
|
540
|
|
|
Michi Henning |
9 years ago
|
|
|