On 28.10.2021 16:11, Rony G. Flatscher wrote:
>
> Tried for some time now to duplicate the error without Java in the picture by 
> duplicating some of
> the code patterns that get applied there in c++ and an ooRexx program.
>
> However so far I was not able to recreate that crash.
>
Today I had again some free time and after editing the test program further I 
let it run for an
hour, however it crashed (after 52,400 Rexx interpreter instances created, 
around minute nine or
ten). (Please note: most of the time the program runs without a crash, 
sometimes it crashes, it is
for the first time that it crashed with a DEBUG 32-bit version from trunk.)

Microsoft Visual Studio is not able to attach:


Looking at the process through Process Explorer shows the following two entries:

It is 32-bit ooRexx on a 64-bit system so 2 GB seems to be stressing some limit 
in the 32-bit process.

Here the rxapi daemon:

---

The "testInterpreterInstances.cpp/exe" tries to mimic what happens in the Java 
server when Rexx
scripts get run, by looping for a given number of seconds (in this case one 
hour was planned, but
the program crashed within ten minutes):

  * a Rexx interpreter instance (rii) gets created,
  * a package gets dynamically required and its package used in that rii to add 
it to other Rexx
    scripts that get created with NewRoutine(),
  * a Rexx routine from the dynamically required package will redirect .input, 
.output, .error,
  * a Rexx routine from the dynamically required package will be called that 
causes with a reply to
    have two calls run on a separate thread:
      o one call causes a native method to run two Rexx programs in the passed 
rii to be created and run
      o the second call will then terminate that particular rii

This was with a DEBUG version of 32-bit ooRexx from trunk ( r12308), in case an 
attachable crash
occurs to be able to see where in ooRexx this happens. (Again, the crash occurs 
rarely, I have not
yet found a combination that would reliably crash, so still work to do.)

One more observation: the longer the program runs (it reports every 100 new 
rii) the creation of rii
becomes noticeably slower.

Not sure, whether this is helpful. If there is anything I can do or supply in 
addition, please let
me know.

---rony


_______________________________________________
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Reply via email to