problem found by robert@ ok jamsek, op

"sed -i" is fundamentally unportable. GNU and OpenBSD sed(1) treat the extension for the backup file as an optional argument and use "sed -i" for no backup file. FreeBSD sed(1) treats the extension as an obligatory argument and uses "sed -i ''" for no backup file. There is no single syntax that works for both. ok stsp op

ok stsp

This behaviour is already documented in got-worktree(5) but wasn't actually implemented. ok stsp@

The idea is that got patch can simply look for a line such as: commit - abcde1234567... to find the merge base commit ID to show in diff3 conflict markers. got log -p now displays commit IDs in its diff header, instead ofl tree or blob IDs. ok op@

This brings the rest of the regression test scripts in line with patch.sh.

This is important in cases like 'got diff file.txt file.txt' which should only show the diff for file.txt once. suggested by kn@

To minimise the amount of churn between the difference in date(1) and ln(n) semantics, use gdate and gln from coreutils.

GNU ln uses -T instead of -h for no-target-directory. OK thomas.adam

ok stsp

Previous default output remains when test cases are run individually. ok tracey

This is a racy solution that needs to be properly implemented in the future. ok stsp

The command could fail with "got: no such entry found in tree". This problem is reproduced by the regression test added in this commit. This happened because file index entries were processed in the wrong order by diff_fileindex_dir(). To fix this, keep removed entries in the RB tree and skip them when the file index is written out, rather than removing entries from the RB tree immediately causing side-effects for RB_NEXT and friends.

and add a regress to check the revert isn't aborted anymore