`PatchSet' API mismatch causing a total `contrib/mklog.py' failure?

Jonathan Wakely jwakely.gcc@gmail.com
Mon Nov 16 12:00:21 GMT 2020


On Mon, 16 Nov 2020 at 11:51, Martin Liška wrote:
>
> On 11/16/20 12:17 PM, Maciej W. Rozycki wrote:
> > On Mon, 16 Nov 2020, Martin Liška wrote:
> >
> >>>    I have decided to give your `contrib/mklog.py' script a hit and, well,
> >>> ahem, I guess I must be doing something utterly silly, but no matter what
> >>> kind of a diff I hand to the script it does not produce anything unless I
> >>> apply a patch like below to it, in which case the output produced is as
> >>> expected.
> >>
> >> Hm, it should not be really needed. See:
> >>
> >> $ git show > 1
> >> $ ipython
> >> $ from unidiff import PatchSet
> >> $ PatchSet(open('1'))
> >> Out[2]: <PatchSet: [<PatchedFile: gcc/config/i386/i386.h>, <PatchedFile:
> >> gcc/doc/invoke.texi>]>
> >
> >   Hmm,
> >
> > $ ipython
> > -bash: ipython: command not found
> > $
> >
> >>>    What's going on here -- has the API of `PatchSet' changed so much at one
> >>> point?  Can we do anything to prevent someone else from tripping over this
> >>> issue?
> >>
> >> Can you please show how do you use the script and what's the output?
> >
> >   Sure, let's pick your recent change:
> >
> > $ git show 2935ff7eb7ac | contrib/mklog.py
> > $
> >
> > Nothing, as I wrote.  With my tweaked version I instead get:
> >
> > $ git show 2935ff7eb7ac | ../mklog.py
> > ChangeLog:
> >
> >       * gcc/testsuite/g++.dg/ubsan/pr61272.C:
> >
> > $
> >
> > which is not perfect as it failed to pick ChangeLog from gcc/testsuite/,
> > but at least it's a starting point.
>
> Hm, I see proper output:
>
> $ git show 2935ff7eb7ac | contrib/mklog.py
> gcc/testsuite/ChangeLog:
>
>         * g++.dg/ubsan/pr61272.C:


FWIW I also get the expected output using Python 3.8.6 and unidiff 0.6.0.


More information about the Gcc mailing list