What did you do before the bug happened? (Steps to reproduce your issue) run the following with a debugger: `git stash push ':(attr:<somevalue>)'` check the resulting pathspec after parse_pathspec() What did you expect to happen? (Expected behavior) the resulting pathspec should contain ':(attr:<somevalue>)'. What happened instead? (Actual behavior) The resulting pathspec is ':(attr,prefix:0)' This is due to parse_pathspec in git-stash being called with PATHSPEC_PREFIX_ORIGIN as a flag. ':(attr:<somevalue>)' gets parsed into ':(attr,prefix:0)' because prefix_magic() https://github.com/git/git/blob/master/pathspec.c#L112 does not take into account that some magic can have values in addition to their magic names specified in the pathspec. What's different between what you expected and what actually happened? Resulting pathspec is ':(attr,prefix:0)' when it should be ':(attr:<somevalue>,prefix:0)' Anything else you want to add: attr pathspec magic is currently blocked for git stash by GUARD_PATHSPEC https://github.com/git/git/blob/master/dir.c#L2176 But, it looks like it can be unblocked, once prefix_magic() is fixed. Also add-interactive also uses PATHSPEC_PREFIX_ORIGIN as a flag and the resulting pathspec from parse_pathspec shows the same bug. But `git add -i ':(attr:<somevalue>'` behaves as expected and I'm not sure why. It may be using different parts of the pathspec struct down the line to filter out files. [Enabled Hooks] commit-msg pre-commit prepare-commit-msg
