hg bisect seems to fail

[Marcin Kasperski]

> The first bad revision is:
> changeset:   43430:4394687b298b
> branch:      stable
> tag:         bad
> user:        Pierre-Yves David <pierre-yves.david at octobus.net>
> date:        Sat Nov 16 20:07:49 2019 +0100
> summary:     pure: use string for exception in the pure version of base85

Considering you faced some string/bytes-related problem under py3,
mayhaps this is the very version which accidentally changed sth what
impacted the problem. Or mayhaps this change caused recompilation of sth
left uncompiled by accident.

> That is nonsense, the wrong branch, and if you checkout the changeset
> before, it does not compile neither.

In heavily merged history bisect travels via various paths, trying to
find sth sensible. 

Not sure what you mean by „changeset before”, but it
can be on another branch which got the same or similar change etc.
In my checkout this changeset's parent is ~70 commits back and sideways
there are commits from other branches:

    hg log -r 4394687b298b
    changeset:   43666:4394687b298b
    branch:      stable
    parent:      43597:856cce0c255c
    user:        Pierre-Yves David <pierre-yves.david at octobus.net>
    date:        Sat Nov 16 20:07:49 2019 +0100
    summary:     pure: use string for exception in the pure version of base85

(note 43666 vs 43597) so if you mayhaps tried 43665 (43429 in your
checkout), it was unrelated commit from another branch.

In general, if you feel sth is really wrong, full transcript of your
session (showing both which changeset you tried, and which decisions you
issued) would probably be needed to analyse the case.