Hi Ben, thanks for the idea. I reverted the skipped test and added your suggested patch. Can you produce a new source build such that this can be tested via the brew flow?
Fritz > Am 04.11.2023 um 21:01 schrieb Ben Pfaff <b...@cs.stanford.edu>: > > It looks to me like Mac OS is returning EOPNOTSUPP when it should > return EILSEQ. Would you mind trying the following patch? > > diff --git a/src/libpspp/i18n.c b/src/libpspp/i18n.c > index 5e9aa0d59c..7697ac614c 100644 > --- a/src/libpspp/i18n.c > +++ b/src/libpspp/i18n.c > @@ -214,6 +215,7 @@ try_recode (struct converter *cvtr, char fallbackchar, > return out - 1 - out_; > > case EILSEQ: > + case EOPNOTSUPP: > if (outbytes == 0) > return -E2BIG; > if (!fallbackchar) > > > On Sat, Nov 4, 2023 at 11:53 AM Friedrich Beckmann > <friedrich.beckm...@gmx.de> wrote: >> >> Hi Ben, >> >> pspp uses the native iconv library from Apple on MacOS. Apple has changed >> that library from MacOS 13 to MacOS 14. As a result the returned error codes >> seem different. That is the reason why the test "convert invalid UTF-8 to >> ISO-8859-1“ fails on MacOS 14. The problem is discussed here: >> >> https://github.com/fredowski/homebrew-pspp/issues/5#issuecomment-1793464558 >> >> I made a patch to disable that test on MacOS such that the regression will >> work again on MacOS 14. I hope until Apple fixes the iconv library. >> >> Cheers >> >> Fritz
