Author terry.reedy
Recipients barry, berker.peksag, durin42, gward, ncoghlan, pitrou, r.david.murray, terry.reedy
Date 2015-04-15.22:22:06
SpamBayes Score -1.0
Marked as misclassified Yes
Message-id <1429136527.17.0.852959284602.issue17445@psf.upfronthosting.co.za>
In-reply-to
Content
I tried the [Create Patch] button.  Two problems: the result is about 90% concerned with other issues; it is not reviewable on Rietveld.  I will unlink it and upload a cut-down version.

Wtiht the test changes suggested by Berker, I agree that it is time to apply this, with whatever decision we make about 3.4.

I am sympathetic to the notion that there is a regression from 2.x.  There is precedent for adding a feature to fix a bug (in difflib, a new parameter for SequenceMatcher, for 2.7 3 (or thereabouts)).  However, doing so was contentious (discussed on pydev) and not meant to be routine.  The bug being fixed had been reported (as I remember) on four separate issues by four people and seconded by other people, so we really wanted the fix in 2.7.

Would the following compromise work for Mercurial?  The patch already adds a new private function _check_types.  For 3.4, also add _diff_bytes as a private function.  Merge both into 3.5.  Create a 3.5 patch that makes _diff_bytes public by renaming it to diff_bytes, adds the new tests, and documents the new feature.  The What's New entry could mention that the function was added privately in 3.4.4.
History
Date User Action Args
2015-04-15 22:22:07terry.reedysetrecipients: + terry.reedy, barry, gward, ncoghlan, pitrou, durin42, r.david.murray, berker.peksag
2015-04-15 22:22:07terry.reedysetmessageid: <1429136527.17.0.852959284602.issue17445@psf.upfronthosting.co.za>
2015-04-15 22:22:07terry.reedylinkissue17445 messages
2015-04-15 22:22:06terry.reedycreate