Don't close already closed pipes #81
Conversation
a9a8e91 added some `close` calls which attempted to close pipes that we already closed previously. This resulted in the error code we read from the failed child being overwritten, resulting in the `process004` test failing.
|
Btw, the changelog entry for 1.4.3 didn't even mention the code change resulting in this regression: http://hackage.haskell.org/package/process-1.4.3.0/changelog |
Make sure stdin, stdout and stderr pipes are closed when exec call fails, e.g. when the executable doesn't exist.
|
Thanks @bgamari. Is it reasonable to hold off on a release to Hackage until you've tested this in GHC? As a side note, it would be really nice to configure Travis/AppVeyor to run the GHC test cases in this repo, I'm just not sure of how to even go about that. |
It passes validation.
It would be nice if we could make that happen but I to am rather unsure of how to go about it. I can think of a few options,
|
a9a8e91 added some
closecalls which attempted to close pipes that we already closed previously. This resulted in the error code we read from the failed child being overwritten, resulting in theprocess004test failing.