Changing the first build to a UTF-8 locale?

Vagrant Cascadian vagrant at
Mon Aug 30 21:08:12 BST 2021

On 2021-08-30, Holger Levsen wrote:
> On Mon, Aug 30, 2021 at 12:16:40PM +0300, Adrian Bunk wrote:
>> there are currently several reproducible-only build failures due to
>> export LANG="C" in the first build.
>> Would it be OK to set LANG to C.UTF-8 or en_US.UTF-8 in the first build 
>> instead?
>> Building under non-UTF-8 locales feels pretty pointless at this time, 
>> and the build failures this causes in reproducible are annoying.
>> The second build is already using a UTF-8 locale.
>> The Debian buildds are already using C.UTF-8 for years.
> this seems sensible to me and I have implemented this change now, thanks 
> for suggesting it.
> (Should some valid rejection comes up for this idea we can always revert it.)

Not a rejection per se, but we could get some of the benefits of both
approaches by only using LANG=C in the unstable and experimental, just
like done for build paths, leaving testing and stable with LANG=C.UTF-8
(or en_US.UTF-8).

It does occasionally find genuine bugs where something builds
successfully with or without UTF-8, and having diffoscope output to
compare that might be helpful on occasion.

C.UTF-8 also isn't in upstream glibc, so while supported in Debian (and
a *different* implementation in Fedora, as I understand it), maybe it
makes some sense to test (in some cases) a non-UTF-8 "C" locale.

Using en_US.UTF-8 to mean "default language" doesn't seem a great
alternative to me, even as a native speaker of ... well... en_US (not
*sure* if I "speak" UTF-8).

So, some more ideas on the subject, though not strongly opinionated
either. :)

live well,
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 227 bytes
Desc: not available
URL: <>

More information about the Reproducible-builds mailing list