MVS, or z/OS as IBM currently call it, has a built-in Unix environment
known as USS (for Unix System Services). It first came out in 1993 as
MVS OpenEdition, and provides the POSIX-y sort of experience you all
know and love, only running as a subsystem of the main OS.

(I intend to add a page going further into this stuff at some
point. Progress is... not quite 'glacial' [1])

To give you an idea, here's a 'screenshot' (t530 is my laptop):

t530$ ssh zos23
LEE@S0W1:~$ uname -a
OS/390 S0W1 26.00 04 1090
LEE@S0W1:~$ ls -la
total 86208
drwxr-xr-x   7 LEE      SYS1        8192 Jan  2 15:05 .
drwxr-xr-x  23 OMVSKERN SYS1       16384 Dec 13 17:48 ..
-rw-------   1 LEE      SYS1       40061 Jan  4 01:48 .bash_history
-rw-r--r--   1 LEE      SYS1        3552 Dec 20 15:57 .bashrc
-rw-rw-rw-   1 LEE      SYS1          71 Dec 14 13:22 .gitconfig
-rw-r--r--   1 LEE      SYS1         852 Dec 19 14:44 .profile
-rw-------   1 LEE      SYS1        3790 Dec 14 12:22 .sh_history
drwx------   2 LEE      SYS1        8192 Dec 14 19:54 .ssh
-rw-------   1 LEE      SYS1       13187 Dec 20 16:37 .tramp_history
drwxr-xr-x   2 LEE      SYS1        8192 Dec 18 22:46 ctest
drwxr-xr-x  15 LEE      SYS1        8192 Jan  2 17:47 dev
-rw-r--r--   1 LEE      SYS1     43023346 Jan  2 15:06 git-2.14.4_b11.190917.tar.gz
-rw-r--r--   1 LEE      SYS1      913602 Dec 15 11:31 ispfinst.exe
drwxr-xr-x   2 LEE      SYS1        8192 Dec 17 12:11 xmi
drwxr-xr-x   6 LEE      SYS1        8192 Dec 14 12:59 zigi
LEE@S0W1:~$

'OS/390' in the uname output is a bit of a hangover from the 31-bit
days. This is running on z/OS 2.3 (latest version is 2.4, but that isn't
installed on the machine I have access to yet). Commands can be sent to
the 'MVS side' via the 'tso' command:

LEE@S0W1:~$ tso "listcat level('lee.zigi')"
listcat level('lee.zigi')
NONVSAM ------- LEE.ZIGI.V1R3.EXEC
    IN-CAT --- USERCAT.STAFF
NONVSAM ------- LEE.ZIGI.V1R3.GPLLIC
    IN-CAT --- USERCAT.STAFF
NONVSAM ------- LEE.ZIGI.V1R3.PANELS
    IN-CAT --- USERCAT.STAFF
NONVSAM ------- LEE.ZIGI.V1R3.README
    IN-CAT --- USERCAT.STAFF
LEE@S0W1:~$

Things like 'cp' etc. can also talk to both sides of the OS eg. you can
cp between USS files/directories and MVS datasets.

(Zigi is an ISPF front-end to 'git' [2])

So, for the last couple of weeks or so, I've been attempting to get
various well-known open-source packages compiled for USS. Stuff I've got
so far includes a handful of Gnu utilities (make, m4, etc) along with
some other miscellany (perl, lua, that sort of thing).

Now, some OSS packages include MVS ports (usually OS type 'os390' or
something), which is great, because those usually build with very little
manual labour. The ones that don't however, generally have one or more
of the following problems:

- Compiler assumptions

There's an ancient port of GCC/glibc here [3], but I haven't used
it. It's a fairly old compiler, built on a fairly old OS (z/OS
1.1). While the backwards-compatibility in MVS is AFAIK unparalleled by
any other OS, I wonder if it's the same for USS? Plus, software
detecting gcc at build time is probably going to expect a much newer
version than this. Also, that 'This is NOT a production-quality
compiler...' bit is somewhat unnerving :)

The upshot of all this is that I'm using the IBM xlc compiler, which is
nice, but means that a large number of unixy compiler flags will need to
be massaged/changed/removed in order for the IBM compiler to even start
looking at your source. How does this compare to your favourite 'gcc'
command-line?

xlc -D_ALL_SOURCE -D_EXT -D_ISOC99_SOURCE -D_OPEN_MSGQ_EXT -D_OPEN_SYS \
-D_POSIX_SOURCE -D_UNIX03_SOURCE -D_UNIX03_THREADS -D_XOPEN_SOURCE=600 \
-D_XOPEN_SOURCE_EXTENDED \
-qlanglvl=extended:extc89:extc99 -qfloat=ieee -qlongname \
-qhaltonmsg=3296 -qxplink -qdll

Then there's the whole '-L' thing, or the completely different flag format
if you're calling the compiler as cc or c89... yeah, fun.

- What's Unix?

Anyone who's ever done any porting work knows this pain. It's like going
back in time to the Unix wars again - hardly anyone includes 'os390'
code, so you're mostly on your own trying to get something to
compile. #ifdef __MVS__ has become a close friend. Any 'configure'
scripts supplied can be extremely hit-and-miss.

- EBCDIC

Anything doing text processing probably needs to be pored over to check
for ASCII-isms (assumptions about sorting, newline characters and so
on.). Good test suites are a godsend here - any package with an
ASCII-centric view will usually generate a whole bunch of test failures.



Anyway, that's what I've been doing recently. My Gnu grep 3.3 tests
finished while I was writing this, looks like I've got some work to do:

SKIP: z-anchor-newline
========================================================================
Testsuite summary for GNU grep 3.3
========================================================================
# TOTAL: 110
# PASS:  42
# SKIP:  56
# XFAIL: 1
# FAIL:  11
# XPASS: 0
# ERROR: 0
========================================================================
See tests/test-suite.log
Please report to [email protected]
========================================================================


[1] gopher://republic.circumlunar.space/1/~leeb/mf/mvs
[2] https://github.com/wizardofzos/zigi
[3] https://www.cozx.com/dpitts/gcc.html

You are viewing proxied material from republic.circumlunar.space. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.