22 lines
1.1 KiB
Text
22 lines
1.1 KiB
Text
|
Spent several days fixing test suite failures on Windows. This started out
|
||
|
|
really annoying; I had to chase back a "NUL" -- the string not the
|
||
|
|
pointer! -- to a indirect dependency that needed an update to work with
|
||
|
|
recent ghc on Windows.
|
||
|
|
|
||
|
|
Then yesterday I fixed most of the other test suite failures on Windows.
|
||
|
|
But, it became clear that the test suite was only testing adjusted unlocked
|
||
|
|
branches on Windows, and was finding non-Windows-specific problems
|
||
|
|
involving them. So, today I added a fifth pass to the test suite, so it
|
||
|
|
will always test adjusted unlocked branches. And fixed all the problems
|
||
|
|
with them that test suite turned up.
|
||
|
|
|
||
|
|
It turned out there was no good way to use `git-annex import`
|
||
|
|
with an adjusted branch. Merging the imported branch into an adjusted
|
||
|
|
branch is likely to result in spurious merge conflicts, and the merged
|
||
|
|
files don't get adjusted. The solution was adding a new way to merge
|
||
|
|
a single branch in the same way that git-annex sync handles merges:
|
||
|
|
`git-annex merge remote/master`
|
||
|
|
|
||
|
|
Sadly, I think there are still a couple of test failures on Windows.
|
||
|
|
(Can't win em all..)
|