Copying untagged text files

[gngrossi]

@IgorTodorovskiIBM

Installed coreutils-9.6.20250319_030539.zos

Regarding the “cp” command…
I copied an untagged ebcdic text file from a directory into /tmp using (cp and cp -p) and the file tagging is set IBM-1047 and tagged as expected. I tested this without and with _ENCODE_FILE_NEW=IBM-1047 and the results are the same.

But...
I copied an untagged ascii text file from a directory into /tmp using (cp and cp -p) and the file is untagged.
I was expecting the file to be set ISO8859-1 and tagged. Right?

thanks

[IgorTodorovskiIBM]

Interesting, I didn't realize we are adding a tag in the first case. That was probably unintended. I recall adding code to copy the original tag to the destination file, but perhaps our untagged codepage heuristic got in the way when it detects the untagged file to be in EBCDIC.

I think case 1 may violate the principle of least surprise, but I can see the benefits. @AnthonyGiorgio @MikeFultonDev @HarithaIBM what do you think?

[HarithaIBM]

Is it autoconversion feature on z/OS, effecting this?

[AnthonyGiorgio]

I vote for preserving existing tags, but adding automatic tags to files can be risky. I'd prefer to have the user explicitly decide to tag the file instead of doing it under the covers.

[gngrossi]

Here's a snippet from one of my scripts for a basic heuristic to tag 1047.
vi seems to tag files as m IBM-1047 T=off

find . -type f -print | xargs file | grep -E "text|empty" | awk '{print $1}'
| sed 's/:$//' | xargs ls -T | grep -E "untagged|^m IBM-1047 T=off" | awk '{print $4}' | xargs chtag -tc IBM-1047 2>/dev/null && RC=0 || RC=4

[AnthonyGiorgio]

We need to ship an autotag utility in some fashion. Either in zopen or ZOAU, but it needs to exist.