Overview
Artifact ID: | 82c7c42d7f99141bb9cc706e20acf2c61a95c5580a5cb1e5a8a3391772c0a8d3 |
---|---|
Ticket: | 270f78ca95b642fbed81ed03ad381d64a0d0f7df
file mkdir: race condition if two workers creates same directory and one worker deletes it immediately |
User & Date: | aspect 2018-07-12 10:21:53 |
Changes
- Change assignee to "sebres"
- Change closer to "nobody"
- Change icomment to:
I'm not convinced by [1c12ee9e45222d6c]. If I'm reading the patch correctly ..
If EEXIST is raised twice, TclFileMakeDirsCmd simply ignores it. The comment above "goto nextPart" suggests something different, so I suspect this isn't intended.
Removing the call to FSStat means we can report success if the target exists and is a file. This is clearly wrong.
Removing the Tcl_ResetResult() call rings alarm bells, but it looks like that was redundant in the first place.
I think lines 287-289 (post-patch) should simply be deleted so the success case can pass through Tcl_FSStat success on line 258.
- Change login to "aspect"
- Change mimetype to "text/x-fossil-wiki"
- Change resolution to "None"
- Change status to "Open"