This series is twofold. The first 4 patches refactor the revision parsing machine so that it's easier to deal with from a parse-opt based parser: [PATCH 1/6] revisions: refactor init_revisions and setup_revisions. 21 files changed, 70 insertions(+), 54 deletions(-) This is the patch I sent before, revisited. [PATCH 2/6] revisions: split the pure option parsing out from parse_revisions. 2 files changed, 273 insertions(+), 308 deletions(-) This patch splits out the revisions _option_ parsing (as opposed to revisions arguments or pseudo arguments like --all, --not, ...) from the revision parser. The patch is huge, but quite straightforward. [PATCH 3/6] revisions: parse_revisions refactor. 1 files changed, 38 insertions(+), 25 deletions(-) This patch reworks parse_revisions so that it works internally as if it was a parse-opt parser. It's equivalent to the previous code, but changes are tricky, and a separate commit is really worth it. [PATCH 4/6] revisions: split handle_revision_args from parse_revisions. 2 files changed, 52 insertions(+), 46 deletions(-) This patch splits out the last bit of parse-revisions (from the previous commit) so that what parses revisions arguments (refs and pseudo arguments like --all/--not/...) can be called independently. The commit is straightforward but moves code around, hence the separate commit from patch 3 so that one isn't lost in the code moves while reviewing the tricky bits from patch 3. The second part is the git-blame migration for real: [PATCH 5/6] git-blame: migrate to incremental parse-option [1/2] 1 files changed, 118 insertions(+), 102 deletions(-) This bit is really alike the proof of concept I saw. Most of the code is stolen from Linus initial patch. It merely removes the git-blame own options and deal with them with a parse-opt parser, in an incremental way. All in all, it's a pretty straighforward patch. [PATCH 6/6] git-blame: migrate to incremental parse-option [2/2] 1 files changed, 40 insertions(+), 92 deletions(-) This is by far the trickiest patch of the series, though it passes the testsuite fine. This patch uses the function from patch 4 to process the revision arguments. We know the dashdash position from the incremental parse-opt, but we have to deal with the "old way" of passing arguments to git-blame. The new code is really less involved than before, because we deal with a filtered argv array where we only have the revisions arguments (no options anymore) and the <path>. -- ·O· Pierre Habouzit ··O madcoder@xxxxxxxxxx OOO http://www.madism.org -- 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
