On Mon, Sep 07, 2026 at 06:05:39PM +0200, Bruno Haible via Bug reports for the GNU Texinfo documentation system wrote: > Hi, > > For two months now, I see this failing test on NetBSD: > > 2026-07-06T12:23:28.6718530Z FAIL: t/formats_encodings.t 317 - > char_latin2_latin2_in_refs converted file_html > 2026-07-20T11:22:41.7303059Z FAIL: t/formats_encodings.t 317 - > char_latin2_latin2_in_refs converted file_html > 2026-07-27T11:48:04.8645448Z FAIL: t/formats_encodings.t 317 - > char_latin2_latin2_in_refs converted file_html > 2026-08-03T11:52:04.7340300Z FAIL: t/formats_encodings.t 317 - > char_latin2_latin2_in_refs converted file_html > 2026-08-13T14:59:34.3030406Z FAIL: t/formats_encodings.t 317 - > char_latin2_latin2_in_refs converted file_html > 2026-08-31T16:23:21.4800704Z FAIL: t/formats_encodings.t 289 - > char_latin2_latin2_in_refs converted file_html > 2026-09-05T17:40:14.0181080Z FAIL: t/formats_encodings.t 317 - > char_latin2_latin2_in_refs converted file_html > 2026-09-07T08:56:06.9972401Z FAIL: t/formats_encodings.t 317 - > char_latin2_latin2_in_refs converted file_html > > Does anyone wants to investigate it, or should I disable the CI on NetBSD? > > Bruno > > > >
Is there any information on why the test fails? The test results should be in a directory called: tta/perl/t/results/formats_encodings/char_latin2_latin2_in_ref - with the reference results under res_html/ and the results from the test run under out_html/. These should contain three files: _0104-_0105-_0118-_0119.html A-a-E-e.html index.html - which should be identical in both directories. Do you have access to these files in this CI setup? If not, we could try building on a NetBSD system somewhere to see if we get the same test failure. This test has been run for more than two months, in fact it appears to have been introduced in 2012 (commit d31f371ba82c9e786083, 2012-01-27 21:05:48 +0000). So it seems likely that it should be possible for this test to be fixed to pass on this NetBSD system. I don't regularly check the results of this CI system (in fact, I would have to look for previous emails to find a web link) and don't know if Patrice does either so we will not know about test failures unless someone reports them. There was a more recent change in May: 2026-05-28 Patrice Dumas <[email protected]> do not use Perl transliterations in tests of C code ... * tta/perl/t/input_files/char_latin2_latin2_in_refs.texi: remove standalone ogonek that is not transliterated the same in C and Perl. but I don't know if that is likely to be responsible or not.
