Notes from September Developer meeting

7 views

Skip to first unread message

Abe Kornelis

unread,

Sep 10, 2025, 5:45:40 PMSep 10

to z390dev...@googlegroups.com, z390

All,

We just had our September meeting.
John Ganci, Larry Simmons, and Abe Kornelis were present.

We discussed the zCobol progress.
Abe is working on importing the NIST test suite.
The work is stalled because he has created a nested class
and feels unsure how to get the details right.

Anthony is not present, so he cannot inform us about
progress on the port of the MVS Cobol compiler.

We then discussed the optable support for z17.
The initial definition of the instructions is complete.
Abe has added a bash script to supplement the
existing bat and groovy scripts for testing opcode tables.
But he fails to get the bash script to work properly.
John offers to have a look into it.

Bob has been working on a floating point issue.
Since he's not present there is no update on progress.

John brought up the issue of non-default file extensions.
Whereas differences in casing make no difference on
a Windows system, they do make a difference on
Linux and macOS systems. This makes test cases
behave differently, depending on the host OS.

Abe has made changes to the build procedure
to adjust the option string being passed to the java compiler
to the java version being used. This was needed
because the current standard option string uses
keywords that Java V8 did not yet support.

John suggests we may want to drop Java V8 support.
It is really old. We've tried to migrate to a newer
version of Java a couple of years back, but we ran
into some issues. From memory, we think it had to do
with security. Since we don't really understand why
those calls to the security package were coded,
we decided to ask Don if he still has some memory of that.
And agreed to address the issue again next month.

Finally, we discussed an issue brought up by Dan Greiner.
Under certain circumstances Complex Relocatable Expressions
generate incorrect object code. John has figured out where
in the code the problem originates, but we have issues
finding the root cause. Also we're worried that suppressing
the apparent symptom might cause unwanted side effects
or bugs down the line - we have no proper test cases
for validating the effects of such a change. Even though
a complete test run with a kludge seems to resolve the
issue, we think we need to better understand the logic
of the program before we make any changes.
Again, help from Don seems the best option to move forward.