Add a new face `magit-section-secondary-heading', which is intended for sub-sections which themselves have children. The children of the "Bisect Log" section are such sections, their heading looks like "git bisect ACTION HASH" and their bodies are hidden by default. It is non-obvious that these sections have a body so we have to provide a hint. Now their headings are bold, just like those of `file' sections, which, as anyone should know, can be expanded.
instead of `magit-list-branch-names'. The commands which use this reader can all benefit from offering more than just branch names for completion.
That was useful when the first line was a header, but now that the header has moved into the `header-line' this is no longer necessary and causes the margin text of the first commit to be removed (for some odd reason only on MS Windows though). Fixes #1951.
Ignore `file-error' originating from `magit-inside-worktree-p' when a buffer visits a yet-to-be-created file located inside a directory that also doesn't exist yet. It would be better to fix `magit-inside-worktree-p' and related functions instead, but not right before a release.
I though I have done this a long time ago, but it seems I only did so for the theme I am personally using (the light variant of solarized).
The intention when setting this to allow value, is to make it faster. However the problem isn't that Git is slow at outputting the log or Magit is slow at parsing the output, the problem is that `git log --graph' has a constant overhead which does not depend on how many commits are being displayed but the total number of commits. Therefore if this number is to low, it becomes more likely that a user has to endure that delay several times until doubling the number of displayed commits until finally all commits which she wants to be listed are indeed listed.
Do this to ensure that the *Warnings* buffer is displayed when users update from 1.4. to 2.1 and still have obsolete packages installed. If there are compile warnings then the *Compile-Log* log buffer might be displayed in the window previously used for the *Warnings* buffer. Re #1933.
Previously the intention was that `ediff-cleanup-mess' would take care of this. That function expects that the control buffer is current when it runs, but our own hook function, which runs firsts, restores the pre-ediff window configuration, making that buffer non-current. So `ediff-cleanup-mess' ended up doing nothing. Reversing the order does not work because our function also expects that the control buffer is current. Also `ediff-cleanup-mess' does things to the file-visiting buffers that we are not interested in. So instead set the local value of `ediff-quit-hook' to a single function, our own cleanup lambda. This means that the global hook value is ignored, including `ediff-cleanup-mess'. `smerge-ediff' does this too. However that function directly calls `ediff-quit-hook' while we call the trimmed down variant `magit-ediff-cleanup-auxiliary-buffers'. Re #1934.