Windows Installer Encoding Issue

[difegue]

Windows Installer Encoding Issue

[Difegue]

Can't believe I forgot to test those again.
The advantage of using a dumb .bat script means fixing this remotely is easy, though.

Try adding chcp 65001 as the first line of the .bat file. If it doesn't work, chcp 932.
I didn't want to force unicode codepage as it's a bit weird on windows, but I guess we'll have to.

Tried both. Dosen't fix the error.
Adding chcp 65001 did fix display of unicode characters inside the prompt however.

[Difegue]

Right. After a bit more searching, it turns out the problem doesn't lie in the windows commandline, but in perl not being aware we're using a windows codepage.

The above below commit fixed the issue on my dev machine (was able to upload/read/delete an archive with Unicode filenames without trouble), I'll update the Windows .zip on the release page to include the hotfix.

Didn't fix the error for me.
[2018-05-03 01:39:39] [Shinobu] [info] Done!
[2018-05-03 01:39:39] [Shinobu] [warn] Deleting ID 51050a66fd9170c94d3fbc9c4f60da8e6715b482 - File D:/Portable/LANraragi/lanraragi/content/(COMIC1��13) [�f�B�G�b�v�H�[ (� �é���[��)] �y�[�p�[�o�b�N�E���C�^�[ACT.00 (�á�¦���¢�������ç���©�½).zip cannot be found.
[2018-05-03 01:39:39] [Shinobu] [info] Archive count (1) has changed since last cached value (0) OR rebuild has been forced (flag value = 0), rebuilding...
[2018-05-03 01:39:34] [Auto-Tagger] [info] Executing enabled plugins on archive with id 51050a66fd9170c94d3fbc9c4f60da8e6715b482.
[2018-05-03 01:39:34] [Shinobu] [info] Adding new file D:/Portable/LANraragi/lanraragi/content/(COMIC1��13) [�f�B�G�b�v�H�[ (� �é���[��)] �y�[�p�[�o�b�N�E���C�^�[ACT.00 (�á�¦���¢�������ç���©�½).zip with ID 51050a66fd9170c94d3fbc9c4f60da8e6715b482

[Difegue]

well shit - maybe I need to specifically scan for the used codepage at runtime and use that instead of the catch-all 65001. What a pain.

Going to need some basic diagnostic info:

• Windows version (I'm testing on 10 which I suspect is much better at handling Unicode)
• Default codepage (you can get that by typing chcp in a command line )

Windows 10 1709
chcp 932

[Difegue]

Right, 932 is Microsoft's version of SHIFT-JIS.
I recall it handling filepaths differently from the others, so that probably explains it.

You can try changing line 8 of lanraragi/lib/LANraragi.pm on your own from cp65001 to cp932 and see if it fixes the problem. If my diagnostic is correct, it should.

I'll add in codepage detection tomorrow - luckily there's a CPAN package for that.

Did not fix the issue.

[Difegue]

¯\_(ツ)_/¯
I'm officially out of ideas.
I'll switch my dev machine to 932 to do some detailed testing, but for now I guess you'll have to rename your files or switch locales.

Eh, no worries. I still got my vagrant install until this starts working.

[Difegue]

I tried a few things on a Shift-JIS Win10 VM, but there's no real easy way to get out of this one that I know of.

(On the bright side, I did port the QuickStarter to a Powershell script while doing my testing, so that'll probably be in 0.5.3. Doesn't fix the problem here, though.)

Our internal Unicode handling is pretty much rock-solid as of now, but the real issue mostly lies in the Windows console itself and how Windows Perl ports have trouble dealing with it.

It mostly works with ASCII locales, but Shift-JIS and a few others introduce changes to the console that throws us off path.
There are seemingly ways* to deal with codepages, but I'm not going to bloat the main codebase with edge cases for them, as the native Windows version still remains secondary** and there's multiple ways to run LRR on those Windows systems, like Vagrant/Docker.

* This would probably only fix the issue for cp932, other weirdass codepages like cyrillic would still glitch up.
** As a matter of fact, Mojolicious dropped Windows support as of last month but it still works so w/e