zsh fails 7 tests on ppc when built with --disable-unicode9 --disable-multibyte

zsh fails 7 tests on ppc when built with --disable-unicode9 --disable-multibyte

[Erhard F.]

[-- Attachment #1: Type: text/plain, Size: 605 bytes --]

Greetings!

I found the testsuite fail the following tests when zsh 5.7.1 is built with --disable-unicode9 --disable-multibyte:
./A03quoting.ztst: test failed.
./B02typeset.ztst: test failed.
./B03print.ztst: test failed.
./D01prompt.ztst: test failed.
./D04parameter.ztst: test failed.
./V09datetime.ztst: test failed.
./V10private.ztst: test failed.

Downstream bugreport: https://bugs.gentoo.org/653704

Attached please find a config.log, build.log, test.log.

Regards,
Erhard

-- 
 PGP-ID: 0x98891295 Fingerprint: 923B 911C 9366 E229 3149 9997 8922 516C 9889 1295
riot.im: @ernsteiswuerfel:matrix.org

[-- Attachment #2: logs.tar.xz --]
[-- Type: application/x-xz, Size: 30284 bytes --]

Re: zsh fails 7 tests on ppc when built with --disable-unicode9 --disable-multibyte

[dana]

On 4 May 2019, at 15:47, Erhard F. <erhard_f@mailbox.org> wrote:
>I found the testsuite fail the following tests when zsh 5.7.1 is built with
>--disable-unicode9 --disable-multibyte:

Most of these failures are the result of a deliberate choice, mentioned in
workers/37304, not to apply fancy quoting when the shell has no multi-byte
support. The affected tests don't account for this difference. (Also, it isn't
documented under e.g. Parameter Expansion Flags)

I'm not sure how much effort is meant to be expended on tests for uncommon
configurations, but we do account for non-multi-byte shells in a few test
files already; i suppose we could do the same for these. If we do, maybe we
should build a multi-byte detection function into ztst rather than copy/paste
it into each individual file. I can look into it later if there are no other
ideas/preferences

dana

Re: zsh fails 7 tests on ppc when built with --disable-unicode9 --disable-multibyte

[Peter Stephenson]

On Mon, 2019-05-06 at 00:06 -0500, dana wrote:
> On 4 May 2019, at 15:47, Erhard F. <erhard_f@mailbox.org> wrote:
> > 
> > I found the testsuite fail the following tests when zsh 5.7.1 is built with
> > --disable-unicode9 --disable-multibyte:
> Most of these failures are the result of a deliberate choice, mentioned in
> workers/37304, not to apply fancy quoting when the shell has no multi-byte
> support. The affected tests don't account for this difference. (Also, it isn't
> documented under e.g. Parameter Expansion Flags)
> 
> I'm not sure how much effort is meant to be expended on tests for uncommon
> configurations, but we do account for non-multi-byte shells in a few test
> files already; i suppose we could do the same for these. If we do, maybe we
> should build a multi-byte detection function into ztst rather than copy/paste
> it into each individual file. I can look into it later if there are no other
> ideas/preferences

Something like that would be very helpful to sanity check this case.

Uncommon configurations aren't necessarily tested for and removing
multibyte support completely isn't something you'd expect people to do
for anything other than very special uses.  Even so, the pre-multibyte
code is a significant set of code in its own right so if we can test it
so much the better.

Cheers
pws