Add config "outputLogEncoding"

[pengbins]

This change try to fix a codepage problem mentioned in some issues, such as #518 , #267 , #245 and #153.
The output messages of cmake build command are garbled when the default system codepage is not utf8.

It is fixed by adding a user config item "outputEncoding" to set the character encoding of messages in the output console.

• How to use

  • Set "outputEncoding" to your system default codepage. (Ctrl+, and search outputEncoding)
    
  • The default value is "utf8". For Chinese user, set it to "GBK".
  • Supported value list can be found here .

• Results

  • Output console used to be unreadable when the default system codepage is not utf8.
    
  • Output console is readable after set "outputEncoding" to GBK.
    

• How it works
  This change introduces a new dependency iconv-lite.
  The console programs emit text encoded with the default system codepage. This change uses iconv-lite to decode those text to utf8 , so they can be displayed correctly on output console. To decode, iconv-lite
  needs to know the default system codepage which is given by user's config "outputEncoding".

[vector-of-bool]

This change is excellent!!

The output codepage issues has been persistently troublesome almost from the very beginning. I've been unable to test and was never sure how to approach this problem.

One think I am wondering is if there is the possibility to auto-detect the code page. I found that Windows exposes a _CODEPAGE environment variable for the current codepage number. I've found a table of these numbers to their names, but I don't think the names that microsoft provides will work with iconv-lite.

Instead of trying to get them all down immediately, could you add a lookup that uses the _CODEPAGE value to default to your GBK when _CODEPAGE is the value you have on your system? As users report this issue I can add entries to this table so that people won't have to manually change their encoding.

[vector-of-bool]

Alternatively, you could try using this table I made from codepage number to names. Maybe iconv just supports it out-of-the box? That would be awesome, and changing the output encoding setting wouldn't ever be required.

code-pages.ts.txt

[pengbins]

I found that Windows exposes a _CODEPAGE environment variable for the current codepage number.

It seems "_CODEPAGE" is not exposed on my Windows 10.

$ > echo %_CODEPAGE%
%_CODEPAGE%

$ >  set | find "CODEPAGE"
// no result

But we can get current active code page by run cmd 'chcp' with no parameter.

$ >chcp /?
Displays or sets the active code page number.

CHCP [nnn]

  nnn   Specifies a code page number.

Type CHCP without a parameter to display the active code page number.

$ >chcp
Active code page: 65001

Maybe we can try to run 'chcp', extract the codepage number, and translate it to encoding name for iconv-lite.

code-pages.ts.txt

PS. Is this table generated from this page.

[vector-of-bool]

It seems "_CODEPAGE" is not exposed on my Windows 10.

That's annoying.

This looks good! I may do some refactors after merging it, but it won't be a big hastle.

Last question: Can you verify that the auto-detection works on your system?

[pengbins]

Last question: Can you verify that the auto-detection works on your system?

Yes, it works on my system. But I didn't test on other language system except mine.
My code page number is 936, encoding name is gb2312.

[MBurtsev]

Problem is still here

[no-realm]

@Nim Can you please open a separate issue for this. It's easier to track that way :)

[MBurtsev]

microsoft/vscode#72270