Junio C Hamano wrote: > Brandon Casey <casey@xxxxxxxxxxxxxxx> writes: > >> Mainly, I prefer to not modify the data structures when a failure occurs. > > Ok. Is the rest of your patch that fixes callers Ok with that > semantics? yes. > If so, I'd agree that is probably cleaner. I'll > scrap the one we are discussing, resurrecting only the api > documentation part, and replace it with the lockfile.c changes > from your patch, along with the fixes to callers. Most of that patch is straight forward, just removing close(). I think you should consider how to handle fdopen on the lock descriptor and the fact that start_command closes the lock file descriptor in create_bundle(). After we fdopen, we should always fclose() and never close(). This isn't enforced. I merely assigned the file descriptor to -1 when it was safe (i.e. after fclose), and added a comment. We could add another function which did this automatically, but maybe that is too much effort, especially in the bundle case. -brandon - To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
