Ludovic Courtès <ludo@gnu.org> writes:

> Hi Marius,
>
> Marius Bakke <mbakke@fastmail.com> skribis:
>
>> Marius Bakke <mbakke@fastmail.com> writes:
>>
>>> Julien Lepiller <julien@lepiller.eu> writes:
>>>
>>>>> Could you send an updated patch to address those issues?  Then we can
>>>>> happily apply it.
>>>> here is the updated patch. I let the tests be done because it doesn't
>>>> harm, but it does no good either, so feel free to disable them if you
>>>> prefer.
>>>
>>> Hello! I've gone through this package and made some improvements. I also
>>> managed to enable the test suite, but had to resort to strace to figure
>>> out what was going on (/bin/sh calls..).
>>
>> Oops, here is a version that actually passes `guix lint`.
>>
>> From 83b2913549784e5d3bafd29641e314b308f7df11 Mon Sep 17 00:00:00 2001
>> From: Julien Lepiller <julien@lepiller.eu>
>> Date: Fri, 11 Nov 2016 15:18:29 +0100
>> Subject: [PATCH] gnu: Add php.
>>
>> * gnu/packages/php.scm: New file.
>> * gnu/local.mk (GNU_SYSTEM_MODULES): Add it.
>
> Thanks for the update!
>
>> +                            ;;"bcmath/libbcmath"
>
> So this one is very different from ‘bc’?

Yes, it's a completely different project and apparently more akin to
'gmp'. It's maintained within the PHP repository.

>> +                            ;;"fileinfo/libmagic" ; This is a patched version of libmagic.
>
> Here adding a dependency on ‘file’ instead (the package that provides
> libmagic) didn’t cut it, right?

Correct. The build system expects to build this, and it appears to be
heavily patched, so I did not try very hard.

>> +         (add-before 'check 'set-test-environment
>> +           (lambda _
>> +             ;; Without this variable, `make test' passes regardless of failures.
>> +             (setenv "REPORT_EXIT_STATUS" "1")
>> +             #t)))
>> +       #:test-target "test"
>> +       #:tests? #t))
>
> Do you have more info as to what fails?  I wouldn’t be surprised if
> there were lots of issues having to do with the use of /bin/sh, host
> name lookups, and other things not available in the build environment.
>
> Could you gather more info as to what fails so that we can either fix
> it, or at least convince ourselves that these are not serious issues and
> document them in a comment?

Some failures are indeed due to missing network or programs in the build
environment. I tried patching a few just now, but unfortunately some
files are in a format apparently not supported by Guile!

 870: 5 [call-with-input-file "ext/mbstring/tests/bug26639.phpt" ...]
In /gnu/store/ciqw5z470c8ihl1kfswj1j3ix6hs092d-module-import/guix/build/utils.scm:
 556: 4 [#<procedure 16a6440 at /gnu/store/ciqw5z470c8ihl1kfswj1j3ix6hs092d-module-import/guix/buil
d/utils.scm:555:10 (in)> #<input: ext/mbstring/tests/bug26639.phpt 11>]
 592: 3 [#<procedure 1798aa0 at /gnu/store/ciqw5z470c8ihl1kfswj1j3ix6hs092d-module-import/guix/buil
d/utils.scm:578:6 (in out)> #<input: ext/mbstring/tests/bug26639.phpt 11> ...]
In ice-9/rdelim.scm:
 188: 2 [read-line #<input: ext/mbstring/tests/bug26639.phpt 11> concat]
In unknown file:
   ?: 1 [%read-line #<input: ext/mbstring/tests/bug26639.phpt 11>]
In ice-9/boot-9.scm:
 109: 0 [#<procedure 16a6480 at ice-9/boot-9.scm:100:6 (thrown-k . args)> decoding-error ...]

ice-9/boot-9.scm:109:20: In procedure #<procedure 16a6480 at ice-9/boot-9.scm:100:6 (thrown-k . arg
s)>:
ice-9/boot-9.scm:109:20: Throw to key `decoding-error' with args `("scm_getc" "input decoding error
" 84 #<input: ext/mbstring/tests/bug26639.phpt 11>)'.

`file` reports: ext/mbstring/tests/bug26639.phpt: Non-ISO extended-ASCII text

Can we work around this somehow? There are more files with non-standard
file types that can't be substituted.

I also found a flag "SKIP_ONLINE_TESTS" which should rule out some more
failures. Attached is an updated patch, but there is still a long way to
go. I will continue this journey tomorrow: