…er is replaced with an underscore: The old file was not getting removed upon publish. Reported by Ashlee Caul (Bug #1446). [David]
document you pass to it is checked out. This is because publishing a story that's checked out can cause no end of havoc. The solution, as mentioned in the error message, is to pass `published_version => 1` when looking for documents to publish from within templates.
…get stuck on a deploy desk.
…ts to make sure that this was really fixed for the two cases I was aware of it. So now it should bloody well stay fixed! A couple of other things: * Using a simple file name, instead of a full path, when naming a file in `t/Bric/Biz/Asset/Business/Media/DevTest.pm`. * Simplifying the test for the presence of XML::LibXML in Bric::Dist::Action::DTDValidate. * Eliminated a level of indentation in `Bric::Util::Burner->_expire()`.
…to be a bug in cpan.pl in trunk.
…onger cause an error.
…ell, all the tests pass. So note that it was done.
…ocuments so that SQL `LIKE` wildcard characters are ignored, so as to avoid confusing, for example, `/foo/bar` with `/foo_bar`. Reported by Scott (Bug # 1368).
…t version of an asset is published while newer versions are being edited. What I've done is prevent the `current_version` column from ever being updated unless the document is currently being checked in. I'm not entirely convinced that this completely fixes the problem, because it only seems to come up when when a new version is created while an older one is in the process of being published. I suppose that it can happen for long publishes, and it definitely is a race condition that I was able to replicate in the tests and then fix, so maybe it will eliminate the problem. Keeping my fingers crossed!
…r template will fail if the version ID specified is the same version number as the currently-checked-out version. This made publishes break, as noted in Bug #1417. I get around it by eliminating the code that checks for whether a version is checked-in or checked-out, the same as the `published_version` parameter does. Tests were added to replicate the issue before I fixed it. Also, I changed the `checkin()` method to increment `current_version` instead of `version`, hoping this will fix the bugs we had with bogus version numbering. I'm going to revert that in a bit and write some more tests to replicate it, then make sure that's really the correct fix. Later tonight, I expect. I also removed the stuff added in r8408, as that was not the correct fix to Bug #1417 (it ''really'' broke the publish lookups!).
…ects the time of the start of a transaction, while the former reflects wallclock time down to microsecond accuracy, meaning that events logged within a single transaction can be better sorted for display.
…blished is currently checked-out to another user (Bug # 1417).
…the UI checkin code. Hoping that this solves the issue with duplicate version constraint violations.
…iles like .JPG won't be matched. Added a corresponding lc() to &$lookup_ext.
… were created (for example, when a story is published to a destination with no associated servers). Reported by Ashlee Caul.