Some help with nroff_output_varies_by_locale_or_utf8

Louis-Philippe Véronneau pollo at debian.org
Tue Sep 17 21:00:16 BST 2019


On 19-09-16 06 h 32, Chris Lamb wrote:
> [Keeping you in CC as requested but no need to keep me in return]
> 
> Hi Pollo,
> 
>> From what I understand, that problem is the
>> "nroff_output_varies_by_locale_or_utf8" one.
> 
> Just a little bit of a heads-up to you and all others reading this
> list: assuming one was looking at a "tagged" package in our notes.git
> repository they not only curated by a (flawed) homo sapiens the
> categorisiation is sometimes in quite some haste. Thus they are really
> not 100% trustworthy -- they are really more intended as a rough guide
> to statistics and what might be worth prioritisation.
> 
> What this means is that one should therefore not rely on them being
> perfect and invest too much time and energy down any particular rabbit
> hole or assuming the problem is at the mentioned or indeed any
> particular abstraction level — it can often be misleading in the diffs
> at a quick glance and thus the package gets miscategorised. I would
> basically just not want you to waste time thinking these were
> canonical…

Hmm, good to know, thanks.

>> I couldn't find any information online on how to tackle this problem.
>> Since my name has an "é" in it and I've written a few manpages already,
>> I fear this will affect other packages.
> 
> … but you seem to have found the rough problem at least.
> 
> In order that I don't duplicate your efforts though, how much
> investigation have you done already? Can we nail it down to rst2man
> specifically? That is to say, can you get rst2man to produce different
> output depending on whatever specific locale variation reprotest is
> introducing? That would be my first step anyway.

This is my first introduction to the wonderful world of
reproducible-builds debugging and I don't have tons of experience with
your tooling :)

So far, I've ran the "atomic-reprotest" Pipeline on Salsa a few times
for rename-flac and I did get it to fail, but on the "num_cpus" test
[1]. I was expecting some failure on the locales test...

Looking at the rst2man code, it also seems it is using:

"locale.setlocale(locale.LC_ALL, '')" [2]

Again, not super experienced here, but wouldn't patching this out or
making it use a consistent locale like C.UTF-8 help solve
reproducibility problems?

I couldn't find anything obviously wrong in manpage.py [3].

Hope that somehow helps.

[1]: https://salsa.debian.org/pollo/atomic-reprotest/pipelines/72814
[2]:
https://salsa.debian.org/python-team/modules/python-docutils/blob/debian/master/tools/rst2man.py#L17
[3]:
https://salsa.debian.org/python-team/modules/python-docutils/blob/debian/master/docutils/writers/manpage.py

-- 
  ⢀⣴⠾⠻⢶⣦⠀
  ⣾⠁⢠⠒⠀⣿⡁  Louis-Philippe Véronneau
  ⢿⡄⠘⠷⠚⠋   pollo at debian.org / veronneau.org
  ⠈⠳⣄

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: OpenPGP digital signature
URL: <http://alioth-lists.debian.net/pipermail/reproducible-builds/attachments/20190917/fd71a191/attachment.sig>


More information about the Reproducible-builds mailing list