On 2020-06-22, Michael Tokarev wrote:
> 22.06.2020 07:26, Vagrant Cascadian wrote:
> ...
>> -RELEASE="$(USER)@$(HOSTNAME) release $(shell cat ../VERSION)"
>> +RELEASE="release $(shell cat ../VERSION)"
>
> Here and in other similar places, how about using RELEASE=...
> explicitly in the parent makefile (which is d/rules) instead
> of patching?

For the most part, whatever works!


> I dunno, I'm just a bit dislike patching and prefer to override
> a variable like this if it's possible.
>
> Patch is good in a way that it works as long as the problem place
> stays the same, - so that if anything there changes (including the
> case when the patch isn't needed anymore), we'll have to revisit
> it at least.

If the patches are upstreamable (many would need reworking), then
getting them upstream would help other distros working on reproducibile
builds, rather than just working around it in debian/rules for Debian.


> Somewhat similar is the situation with LC_* variables, - sometimes
> I just export LC_ALL=C in the top of d/rules and be done with it,
> after that I don't care how well the upstream build system deals
> with various intl. stuff. Note that sometimes non-C locale might
> expose difficult to debug bugs, - for example, in the C locale
> the order of upper- and lower-case letters (first all uppercase
> next all lowercase) is different than in some other locales,
> like in RU, where upper and lower-case is intermixed (first Aa,
> next Bb etc), - so when doing thigs like $(ls -1 | head ..)
> we might have entirely different results.
>
> What do you think?

Other than the potential for upstreaming, that sounds good too.

If just going the debian/rules route, I'd recommend using
LC_ALL=C.UTF-8, although it's not in upstream glibc, it's always
available in Debian and gets you UTF-8 out of the box.

Though I haven't tested on all locales, the sort order seems to be more
consistant across UTF-8 locales, though haven't tested all locales.


> And thank you for debugging all this!

Thanks for the quick response!


It's been a bit different building on a 32-core machine with ~150GB of
ram than my old dual-core with 4GB of ram back when I used to maintain
qemu... :)


live well,
  vagrant

Attachment: signature.asc
Description: PGP signature

Reply via email to