[Debian GNUstep maintainers] Bug#888908: using ICU 60.2: Failed test: basic10_4.m:166 ... negativeFormat used for -ve number

[László Böszörményi (GCS)]

On Thu, Feb 1, 2018 at 3:45 PM, Yavor Doganov <yavor at gnu.org> wrote:
> Matthias Klose wrote:
>> On 31.01.2018 08:51, László Böszörményi (GCS) wrote:
>> >  Strange, I'm at least twice built the ICU reverse dependencies
>> > successfully including gnustep-base.
>
> That would have been with versions before 59.1-1.  The configure run
> fails because the icu-config script has been removed.  You have to
> pass --disable-icu-config to build gnustep-base.
 Exactly, I did that and it built with ICU 60.2-3. Yesterday when I
tried to build again the same gnustep-base source (I still have it as
a local NMU with the disabled icu-config) the mentioned test failed.
Hence I think it's not ICU itself which changes the 'negativeFormat'.
I've the buildinfo file from the previous successful compilation and
will check what's changed.

> What are the plans for the ICU transition?
 I want to rebuild all of the reverse dependencies at least on amd64
(but better on i386 and/or arm64 as well). I'm a bit stressed and my
progress is much slower. :( Partial result shows all issues can be
easily patched. The only problem at the moment is that I was slow
enough and some Haskell metapackage changed that breaks higher level
Haskell packages. I think it's better to restart my test buildings and
do it on one day if possible. Also check the gnustep-base self-test
failure as noted previously I could build the package with the latest
ICU from experimental. Something else changed since then.
In short, I hope I can ask the RMs about it next week.

Regards,
Laszlo/GCS