Use UTF-8 encoding for @cmdfile

[dlangBugzillaToGithub]

Dāvis (@davispuh) reported this on 2015-04-20T21:53:18Z

Transferred from https://issues.dlang.org/show_bug.cgi?id=14474

CC List

• Martin Nowak (@MartinNowak)
• godmyoh

Description

Currently `dmd @cmdfile` will read file as simple char string. But when DUB is compiling and creating this file it always saves it in UTF-8.

This makes a problem when cmdfile contains paths (source files) with non-ASCII paths/names and thus DMD interprets them incorrectly (it always uses WinAPI ANSI functions which expects them to be encoded in Windows default ANSI code page)

Such cmdfile file with non-ASCII paths encoded as UTF-8 will produce "Error: cannot read file"

One solution could be to change DUB to save it in Windows default ANSI code page, but I think that's a very bad idea because then that file won't be portable.

So best would be to enforce it being in UTF-8 encoding and then decode in DMD for respective code page which is used for WinAPI calls.

[dlangBugzillaToGithub]

davispuh commented on 2015-04-20T22:21:13Z

Created a quick PR https://github.com/D-Programming-Language/dmd/pull/4602

[dlangBugzillaToGithub]

code (@MartinNowak) commented on 2015-04-22T00:08:17Z

(In reply to Dāvis from comment #0)
> One solution could be to change DUB to save it in Windows default ANSI code
> page, but I think that's a very bad idea because then that file won't be
> portable.

It doesn't need to be portable, it's just used for a single compiler invocation.

> So best would be to enforce it being in UTF-8 encoding and then decode in
> DMD for respective code page which is used for WinAPI calls.

UTF-8 sounds reasonable.

[dlangBugzillaToGithub]

menodinulla commented on 2019-11-02T08:30:30Z

Is anyone working on this issue?

This PR is closed but doesn't seem to be resolved.
https://github.com/dlang/dmd/pull/4602