[Reproducible-builds] Bug#815171: Bug#815171: Bug#815171: Bug#815171: diffoscope: build time tests fail on armhf

[Holger Levsen]

Hi,

On Montag, 22. Februar 2016, Jérémy Bobbio wrote:
> Because it's just a test that is brittle. It doesn't 
affect normal use
> of the installed package and does only prevent a 
successful build one
> times out of ten.

only counting version 47+48, it caused 5 build failures 
out of 5 build attempts on reproducible.debian.net on 
armhf and 3 build failures out of 9 build attempts on 
amd64, thus I will upgrade the severity of this bug to 
serious again.
 
(I havent checked closely whether due to the same issue 
but versions 45+46 also showed random build successes and 
failures on armhf…)

https://tests.reproducible-
builds.org/history/diffoscope.html is where I got this 
numbers from.

> To please the cruel god of FTBFS I can also disable 
running the test
> suite entirely at build time. 

WTF. I'm not working here to please the "cruel gods of 
reproducibility" but because I think it's a worthwhile 
goal. Similar, FTBFS is a serious issue.

Did you really have to ridicule my argument? :-(

"The cruel gods of FTBFS" might be a funny expression in 
general, but it is not, if you are being told your 
bugreport is silly. Think about it.

> I believe this would be the wrong thing to
> do, but if you insist, I'll just do that.

No, what I initially wanted (well, after your first 
reply…) was that this bug was _not_ treated as a "normal" 
one, but as an "important" one, that's all. 

(And to me that's an important disctinction. I read the 
downgrade to "normal" as "bah, I dont care" while I would 
have read "important" as "sure, we will fix this" but it's 
not that serious as the package works fine and we're still 
in the development phase of stretch".)

But given the above (100% build failure on armhf and 33% 
failure on am64) I now think this is indeed a serious 
issue, thus it should be treated as such. 

If your only solution is to disable the test, so be it. 
I'd rather not have this test than having this package 
unreproducibly failing to build.

We cannot^wmust not point fingers^w^wfile serious bugs at 
other packages while having other standards for our 
packages.


cheers,
	Holger
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 828 bytes
Desc: This is a digitally signed message part.
URL: <http://lists.alioth.debian.org/pipermail/reproducible-builds/attachments/20160223/16d2edcf/attachment.sig>