Handling of mixed-codepage files in git

[stevchen-ibm]

We have some mixed-codepage files that are in mostly ibm-1047 but mixed with other codepage such as Japanese.
We can use -text in .gitattributes to check in the files, however using this approach we will lose the ability to do blame/diff the part that are in text (ibm-1047 code).
Perhaps a solution will be to support zos-working-tree-encoding=ibm-1047 -text in .gitattributes?

[nickrayjones]

I'm having a similar issue with -text. It appears that using -text in .gitattributes results in a binary tagged file on z/OS.

I recently discovered this when .gitattributes has:
* -text git-encoding=utf-8 zos-working-tree-encoding=ibm-1047 working-tree-encoding=utf-8

Normally we have this set
* text git-encoding=utf-8 zos-working-tree-encoding=ibm-1047 working-tree-encoding=utf-8
And the file is tagged as text with the expected code page ibm-1047 on z/OS

While not intuitive, that the meaning of -text is not 'load this file as binary'. -- https://git-scm.com/docs/gitattributes#_effects
The -text should mean

Unsetting the text attribute on a path tells Git not to attempt any end-of-line conversion upon checkin or checkout.

I think -text could be honored by using the chattr command to set the desired flags (along with chtag).

To get files as binary the suggestion is to unset the diff attribute (-diff) but the binary attribute was later added to mean -text -diff
https://git-scm.com/docs/gitattributes#_marking_files_as_binary