Coordination with upstream PCRE2

[NWilson]

Hi,

I'm the maintainer of the upstream PCRE2 project.

We are currently discussing whether PCRE2 is able to support, or should support, IBM z/OS.

We don't have access to a z/OS machine (nor do we really want to spend much of our time on this issue).

Currently, PCRE2 supports EBCDIC, which is great, but we wonder if anyone at all is actually even using this.

With a PR which is open, all our tests will pass under EBCDIC (PCRE2Project/pcre2#656). Currently the tests are broken for EBCDIC, and I believe it has been buggy for some time. What is worrying is that no-one has reported this.

1. Does your zopen port of PCRE2 use EBCDIC, or are you compiling for an ASCII environment?
2. I also would appreciate knowing more about the different operating systems used on IBM mainframes. When we talk about "S390X" as a CPU architecture, I presume that can run more than one OS, and some of them are ASCII and some are EBCDIC? Does "z/OS" mean "EBCDIC" or can that be ASCII these days?
3. Would you be upset if PCRE2 dropped support for EBCDIC encoding? Do you know of other PCRE2 users on IBM hardware who might be using EBCDIC still? PCRE2 is used inside other software such as PHP, R, and Julia.
4. A previous contributor to PCRE2 (Ze'ev Atlas - his website) added support for "native z/OS without Unix System Services". Is this something which is common these days? I believe that "Unix System Services" would be the normal way to compile & use PCRE2 - and that USS is also to support EBCDIC (which is why IBM insists that POSIX does not specify that a "UNIX" environment is ASCII).

[IgorTodorovskiIBM]

Hi, for zopen, we are building pcre2 and all other open source software in ASCII/UTF8 mode. z/OS is BI-MODAL now and supports both ASCII and EBCDIC. It uses file tags to specify the codepage for a given file. The underlying C runtime services then perform an auto-conversion from ebcdic tagged files to ASCII.

Due to the increased effort of maintaining open source software for EBCDIC builds, we've abandoned this path and simply build everything in ASCII/UTF8 mode.

z/OS and Linux on Z both are s390x. For z/OS we use the __MVS__ platform macro for anything z/OS specific.

If you'd like to reach a broader IBM Z audience, I would suggest asking in the System Z discord: https://discord.com/invite/sze

[NWilson]

Thank you very much Igor, that's helpful to know.