Issue tracker cleanup: SVN-2079

[Nathan Hartman <ha...@gmail.com>]

For this week's issue, SVN-2079 "utf8_tests.py should be made non-
iso8859-1 specific":

According to the issue tracker, the utf8_tests.py fail on non-iso8859-
1 systems and it seems that much smarter people than myself were
stumped as to how to fix it.

* This issue was reported in 2004.
* In 2005, Ben Collins-Sussman disabled the test (mark 'Skip').
* In 2013, Bert Huijben moved to the legacy subdirectory.
* It still lives there today.
* It appears the test has not been run since 2005.

Some questions:

How important is it to run this test?

Has this test been superseded by something else over the years?

Have there been new developments in Python that would help alleviate
whatever stumped everyone in the past?

Should the test be removed?

Thanks,
Nathan

Re: Issue tracker cleanup: SVN-2079

[Nathan Hartman <ha...@gmail.com>]

On Sun, Nov 3, 2019 at 4:35 PM Branko Čibej <br...@apache.org> wrote:

> On 03.11.2019 17:45, Nathan Hartman wrote:
> > For this week's issue, SVN-2079 "utf8_tests.py should be made non-
> > iso8859-1 specific":
>
> That test has two problems: first, it depends on the locale where it's
> being run; and second, it's very Unix-centric. We can't really fix the
> first part because the whole point of the test is to verify that
> Subversion correctly converts its (command-line) input to UTF-8. Then,
> if we somehow manage to work around the locale-dependent part, I
> wouldn't be surprised if the Unix-centric part would stop being an issue.
>

This sounds like it will take a bit of research...

I think that probably the best thing to do is to reach out to a
community that has expertise in this area.

Because I'm sure that someone, somewhere, has to test utf8
conversions...

Re: Issue tracker cleanup: SVN-2079

[Branko Čibej <br...@apache.org>]

On 03.11.2019 17:45, Nathan Hartman wrote:
> For this week's issue, SVN-2079 "utf8_tests.py should be made non-
> iso8859-1 specific":
>
> According to the issue tracker, the utf8_tests.py fail on non-iso8859-
> 1 systems and it seems that much smarter people than myself were
> stumped as to how to fix it.
>
> * This issue was reported in 2004.
> * In 2005, Ben Collins-Sussman disabled the test (mark 'Skip').
> * In 2013, Bert Huijben moved to the legacy subdirectory.
> * It still lives there today.
> * It appears the test has not been run since 2005.
>
> Some questions:
>
> How important is it to run this test?

That test has two problems: first, it depends on the locale where it's
being run; and second, it's very Unix-centric. We can't really fix the
first part because the whole point of the test is to verify that
Subversion correctly converts its (command-line) input to UTF-8. Then,
if we somehow manage to work around the locale-dependent part, I
wouldn't be surprised if the Unix-centric part would stop being an issue.

I think it's quite important to test the encoding conversions.

> Has this test been superseded by something else over the years?

Not that I can remember.

> Have there been new developments in Python that would help alleviate
> whatever stumped everyone in the past?

Maybe there's something in Python 3 that we could use to generate the
input data in a platform-agnostic way. Python 2 had far too rudimentary
support for different encodings at the time.

> Should the test be removed?

No.

-- Brane