Denton Liu <liu.den...@gmail.com> writes: > For yours, I would reword your commit message to something like > > t8014: avoid git command in upstream pipe > > Use an intermediate file between between git blame and sed to avoid > git blame's exit code being hidden.
I agree that the main "points" of this patch that should be highlighted on the title line are that it is about a test, and it is about not hiding a failure of a Git command by placing it on the upstream side of a pipe---the above title is very nicely written. >> + "git blame -M9 --ignore-rev $IGNOREME $i >output && >> + sed -e \"$pick_author\" <output >actual && > > We should take advantage of the fact that sed can open its own input > here. So we should drop the `<` and just pass the filename to sed. Same > applies to the below. While I do not think it matters too much in this case, I agree it is a good habit to get into, because it would give the command a chance to produce a better error diagnosis (i.e. "malformed input on line X" vs "malformed input on line X in file F"), when it wants to report an error in input, if we give the name of the file to open to the command instead of an already-open file descriptor, for one thing.
