On Tue, May 13, 2014 at 7:57 AM, Paolo Bonzini <[email protected]> wrote: > Il 13/05/2014 16:50, Jim Meyering ha scritto: > >>> > Well, a patch for that was posted more than 1 month ago. >> >> If you mean the trivial revert, I had the impression we'd >> discussed that already. > > > I mean the trivial revert. Regressions should always trump code cleanup.
Sure, but that was not pure "cleanup". It removed code that was error prone, and that had already resulted in one bug due to misunderstanding of how it worked. I suppose by "regressions" you are referring to the risk that some reasonable portability target lacks a shell, and that by relying on shell-based fgrep/egrep wrappers, we are inducing a regression on that system? If someone finds such a system, I'm sure we'll fix it in the next release, but I won't see that a regression before I see details of an actual failure.
