Re: [PATCH v2 1/8] tests: Make sure LANG is set properly for iso9660_test

["Thomas Schmitt" <scdbackup@gmx.net>]

Hi,

i wrote:
> >   xorriso ... -as mkisofs -input-charset UTF-8 -output-charset UTF-8

Glenn Washburn wrote:
> Do there need to be any UTF-8 locales installed (or any locales for
> that matter) for this to work? My guess is no.

I expect the same. It's not easy to test, though.

In theory it is essentially a matter of iconv_open(3) and iconv(3) whether
conversion works.
The installed locale only gives the default values for parameters of
  iconv_t iconv_open(const char *tocode, const char *fromcode);
The options -input-charset and -output-charset override the locale
setting, whatever it is.
(Actually xorriso stumbled over the hardcoded output character set of
Joliet, which is UCS-2.)


> Correct me if I'm wrong, but this logic about defaulting LANG is more
> general than the iso9660_test. Would you agree that this should be
> applied to the tests a a whole?

The comments in grub-fs-tester talk of UTF-8 as expected character set
and the text constants look like UTF-8, too. So it would not be wrong
to enforce a UTF-8 locale.

It might be peculiar to xorriso to look at the locale when interpreting
the meaning of bytes which represent characters.
But this interpretation is inevitable if the target filesystem specs
prescribe a particular character set, as Joliet does, and if we aim for
dealing with UTF-8 names.


> If yes, then I should probably remove this patch altogether and put
> this code elsewhere.

I think it can be justified to hardcode UTF-8 in grub-fs-tester regardless
whether higher levels of the test empire or the user have own interests
in locale settings.
The code in grub-fs-tester is a good example for the situation that the
file names on disk might belong to users with differing locales. We expect
UTF-8 to be the common base of all those nowadays.

So i'd vote for something like

  if [ -n "${LANG##*UTF*}" ]; then
     echo "NOTE: LANG=$LANG appears to not be UTF-8."
     echo "NOTE: Setting LANG=en_US.UTF-8 to match the test file names."
     export LANG=en_US.UTF-8
  fi

at the start of grub-fs-tester.in.


> > How can i force "make check" to do something at its second run ?

> However, "make check" should still be running the test.

I obviously got misled by the many make messages and the silence of the
iso9660_test run.


> Would you submit a patch with the changes you've
> outlined that would fix this issue for when LANG is empty?

I'd go farer and smack down everything that does not claim to be UTF-8.

Is a git diff enough or has it to be git format-patch from a local commit ?
Is git send-email necessary ?
(My relation to git based interaction with projects is best illustrated by
  https://upload.wikimedia.org/wikipedia/commons/thumb/1/17/Laocoon_Pio-Clementino_Inv1059-1064-1067.jpg/564px-Laocoon_Pio-Clementino_Inv1059-1064-1067.jpg
)


Have a nice day :)

Thomas