Bugfix for issue#102 -- "5.11 breaks HTML::FormatExternal"

[Perlbotics]

Test case added.
Skipping attempt to unescape empty authority part.

See: #102

[Perlbotics]

Hi Olaf (@oalders), seems that an entry is missing in Changes?
Can I edit this file myself or is that task performed by a tool like dzil somewhere upstream?

[oalders]

That can be done manually so feel free to add the entry. 😀

[Perlbotics]

Thank you, Olaf (@oalders). Changes updated and pushed.

[oalders]

This could also be

Regression test for #102.

[Perlbotics]

It should be added, that I was able to validate this test on Linux only.
It can be added to the generic test data structure in the beginning of t/file.t
if the results for Windows and MacOs have been verified.

[Perlbotics]

Hi Olaf (@oalders), I identified schemes that do not have an authority part and thus do not need a special treatment
of the host part (768e8e6). Early return for these. Not quite necessary, but I think the intention is cleaner now.

The second patch (84176ec) is more important. 'mailto:' has a complex format that needs special treatment.
Since I have no solution for now, the previous 5.10 encoding is restored.
I also added test cases to t/mailto.t that would have helped to detect this bug earlier.

Please consider to publish the corrections or to revert to 5.10 since the current bug (file:///) might
have security implications or could cause havoc on the file system.

Update: regression tests in t/file.t were only verified for Linux. See #103 (comment)

[oalders]

Thanks for this @Perlbotics! Just a few stylistic changes requested. Thank you for being thorough with the added tests.

[simbabque]

Looks good to me. Nice work, thank you.

[oalders]

Thanks for being a good sport through all of these iterations @Perlbotics! I've merged this manually into master as 725fbfb.

Closes #102.

[Perlbotics]

Thank you Olaf (@oalders), Julien (@simbabque), and Graham (@haarg) for your patience and support.
This was the first time, I participated in a github patch activity and while finding great support here, I also learned a lot.

BTW: If you want to improve test coverage, try to run everything once more with something like:
URI_HAS_RESERVED_SQUARE_BRACKETS=1 prove -l
I've done this manually here w/o problems but perhaps it is worth an extra author test?

Thanks.

[oalders]

This was the first time, I participated in a github patch activity and while finding great support here, I also learned a lot.

Congratulations! 🎊 I'm glad it was a positive experience.

If you want to improve test coverage, try to run everything once more with something like:
URI_HAS_RESERVED_SQUARE_BRACKETS=1 prove -l

We can do this in the GitHub workflow, which is at https://github.com/libwww-perl/URI/blob/master/.github/workflows/dzil-build-and-test.yml

Maybe add a job which needs: build-job and just runs the tests again on Linux using Perl 5.36. I don't think we necessarily want to run the test suite twice for the entire matrix. If we just add that one extra run, I think that would be good enough.

[simbabque]

Maybe add a job which needs: build-job and just runs the tests again on Linux using Perl 5.36. I don't think we necessarily want to run the test suite twice for the entire matrix. If we just add that one extra run, I think that would be good enough.

@Perlbotics if you need help with this, give us a shout. I'm happy to explain how these actions work. :)