[Oorexx-devel] FindClass in ThreadContext

[Erich Steinböck]

When a native function package like ooDialog has call-backs which re-attach
through AttachThread() the call-back has just a ThreadContext available.
This makes it difficult for the call-back to search for its own classes
loaded with a ::requires.

FindClass() is available in a ThreadContext but will only search the REXX
package.
FindContextClass() is unavailable for a ThreadContext.
FindPackageClass(), GetPackagePublicClasses() or similar require a Package
object to search.
ooDialog used to re-load the full ooDialog.cls with LoadPackage which I
think is a bit desperate.

Is there a decent way to access own classes from a ThreadContext?
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Discuss bug 1647 - ooDialog crashes interpreter

[Erich Steinböck]

Hi Jon,

> the bug ticket is 'ooDialog 4.2.0' - would this bug get more attention if
> the milestone was '5.1.0'
>
No. The ooDialog.xxx Bug groups are outdated since ooDialog is now a part
of ooRexx 5.

Even if we don't know how to fix this problem, could we not handle it such
> that it throws a rexx error rather than crashing the interpreter?

No, if the problem isn't fixed, it will crash.

Is anyone likely to fix this problem or handle the errors in the (near)
> future?

I've posted a work-around in the bug report at
https://sourceforge.net/p/oorexx/bugs/1647/
Please use it until the issue can be identified and fixed.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Update on ooRexx -> Android

[Erich Steinböck]

Hi Thomas,
the CMake config issues seem to be fixed. Found and not found configs look
reasonable.

I'm not aware of any other ooRexx cross-compile attempt, so this is new
territory for us.
To complete your build you can try commenting out the six lines after #
Build the rexx.img file
and add a line
message(STATUS "rexximage ${ORX_IMAGE_OUTPUT_LOCATION}/rexx.img")
When the build has completed you can run the built rexximage executable in
an Android shell.
The above message will show you which argument you should supply (although
the path might be different between what your compile host sees and what
the Android shell sees).

Moving an existing rexx.img might also work as long as host and Android are
the same bitness and architecture.

Are you cross-building for a different architecture, like from x86 to ARM?
Does the cross build allow building Android 32 bit (when your host is 64
bit)?
Do you run the Android code on an Emulator or are you pushing the build
artifacts to an actual Android hardware?
Did you note down the steps necessary to successfully set up Android NDK
and Emulator for an ooRexx cross-compile?  If yes, can you share the them?
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Update on ooRexx -> Android

[Erich Steinböck]

Hi Thomas,
I've committed a fix to our CMake - it should now be able handle the
missing libpthread on Android.

The attached CMakeError.log don't show all HAVE_PTHREAD_* configuration
errors - not sure why, but it shows issues with libpthread not found which
should now go away.
Please delete the CMake cache, re-run cmake, and provide the CMake build
outout and also again the log file.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Update on ooRexx -> Android

[Erich Steinböck]

Hi Thomas,
it seems that Android has no pthreads lib (it is built-in into libc
instead).  I will make appropriate changes to our CMake.
Can you attach CMakeFiles/CMakeError.log so that we can see why
pthread_attr_get_np isn't found (and also none ogf the others).
Thanks!
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Update on ooRexx -> Android

[Erich Steinböck]

Hi Thomas,
thanks for reporting these issues!

PTHREAD cannot be handled by Android
>
The getStackSize and getStackBase functions require pthread_getattr_np.
It looks like this should be available on Android in  any API level - see
https://cs.android.com/android/platform/superproject/+/master:bionic/libc/include/pthread.h
which lists pthread_getattr_np()
cmake checks for the availability of this function and sets
HAVE_PTHREAD_ATTR_GET_NP appropriately.
Can you please provide the cmake command syntax you are using for the
cross-compile?
Can you please copy/paste the HAVE_PTHREAD_... results that cmake reports?

Please provide compiler error messages whenever possible (copy/paste text
preferred over screenshots).
"caused problems during compilation" isn't enough to investigate

I've just committed fixes for the SO_USELOOPBACK and the struct fd_set
issues.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Help Known compiling error / Linux

[Erich Steinböck]

Hi Thomas,
good to hear that re-running CMake solved the issue.

|  On the ARM Macbook I compiled for the Target System "arm64_v8a". It was
not possible to execute the rexx executable which was build for the target
system arm64_v8a on my macbook with the M2 ARM Chip. That made a lot of
sense to me.
I'm not sure why this should make sense.
A quick check on stackoverflow suggests otherwise, e. g. "Android Studio on
my M1 Macbook Pro can simulate all the system images with ABI arm64-v8a
really well"

|  Now I build on a x86_64 Linux machine for the target x86_64 so it makes
sense that the rexx executable which I build for x86_64 is working on my
source system which is also x86_64. Would you agree here? Im kind of lost.
I'd agree.  When you build for an architecture while running on the same
architecture, the built executable should run.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Cmake deprecation warning

[Erich Steinböck]

Hi P.O.,
this is just a deprecation warning. As soon as we use cmake_policy(VERSION
..) we run deprecated – that's to be expected.
My Ubuntu 16 Jenkins client runs CMake 3.5.1.  I wouldn't bump cmake
minimum until absolutely required.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Help Known compiling error / Linux

[Erich Steinböck]

Hi Thomas,
we successfully build for Ubuntu 22 - you may just have to re-run cmake
again.
Please check which defines (see below) you see as "found" on your system an
report back.
-- Looking for pthread_getattr_np
-- Looking for pthread_getattr_np - found
-- Looking for pthread_attr_get_np
-- Looking for pthread_attr_get_np - not found
-- Looking for pthread_stackseg_np
-- Looking for pthread_stackseg_np - not found
-- Looking for pthread_attr_getstack
-- Looking for pthread_attr_getstack - found
-- Looking for pthread_get_stacksize_np
-- Looking for pthread_get_stacksize_np - not found
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Help Known compiling error / Linux

[Erich Steinböck]

Hi Thomas,
which code are you compiling? The latest trunk?
Which compiler error are you seeing?
Which Linus platform are you on?
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Mark's M1 Mac running macOS 13 Ventura failing two tests

[Erich Steinböck]

Nothing to worry about.
The search_order*testGroup files are temp files created by the REQUIRES
test group and are not intended to be run by testOORexx as actual test
groups.
When not cleaned up properly, they will remain and show this issue, but
will be cleaned up automatically by the next run.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] oorexx net.imap.cls location - retrieve emails

[Erich Steinböck]

Hi Sandy,
we have an SMTP and a MIME class for ooRexx (see rxsock.pdf) but I'm not
aware of an IMAP class.

You could use curl to retrieve Gmail mails, to test try something like this:
curl --verbose -u firstname.lastname -X "search flagged" imaps://
imap.gmail.com:993/inbox
(you'll need a Google "App Password" for this - see
https://support.google.com/accounts/answer/185833)

To automate with ooRexx you could use Rexx/CURL
(note that for the Windows 64-bit platform Rexx/CURL is broken - see
https://sourceforge.net/p/rexxcurl/bugs/14/
I've built a working Windows 64-Bit version in our sandbox - see
https://sourceforge.net/p/oorexx/code-0/HEAD/tree/sandbox/erich/rexxcurl/)


On Sun, Jul 16, 2023 at 12:52 PM Sanford Geiger 
wrote:

> Actually, what I am trying to do is retrieve emails and attachments from a
> gmail server, also from a Hotmail server. Can you help?
>
>
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Unexpected behaviour with serial reuse of address ... with output replace stream 'streamName'

[Erich Steinböck]

Hi Leslie,
you'll want to close the Stream after reading it.
lines() will implicitly open the Stream for r/w and the ADDRESS will raise
NOTREADY:13 on Windows but isn't detected on Linux.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] my experience with Open Object Rexx 5.1 and OpenSuse Leap 15.4

[Erich Steinböck]

Hi Franz,
your address "" "rexxc" works in your MyRexxComp.rex and prints:
all went well with: MyRexxComp.rex , MyRexxComp.cmp.rex has been created

Please provide a piece of failing code and the exact error message you're
seeing.
Note that on Linux (in contrast to Windows) the case of filenames is
significant.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] my experience with Open Object Rexx 5.1 and OpenSuse Leap 15.4

[Erich Steinböck]

Hi Franz,
rexxc is expected to work the same on Windows and Linux.
A minor known difference is that on Windows rexxc will accept both /s and
-s, on Unix only -s is accepted. For best portability always use -s if
required.

Your MyRexxComp.rex should work as long as the rexx filenames in
MyRexxComp.ini are specified without any UTF-8 byte-order-mark (BOM, 'EF BB
BF'x).
The attached MyRexxComp.ini has a BOM in its first line.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] News on Raspberry Pi builds

[Erich Steinböck]

Hi P.O.,
yes, thanks, rebuild the broken installers.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

[Oorexx-devel] GUARD issue

[Erich Steinböck]

If in this, heavily GUARDED example m2 can start running while m1 is
running, and later
m1 can get back control even though m2 is running GUARDED, then how can any
method make sure it runs with exclusive access?

Am I missing something?

~~~
c = .c~new
c~m1
c~m2

::class c

::method m1 guarded
  expose a
  guard on
  reply
  guard on
  say "m1 start"
  guard on when a = 3
  say "m1 end"

::method m2 guarded
  expose a
  guard on
  say 'm2 start'
  do a = 1 to 5
call SysSleep 0.2
say "m2, a =" a
  end
  say "m2 end"
~~~

gives

~~~
m1 start
m2 start
m2, a = 1
m2, a = 2
m1 end
m2, a = 3
m2, a = 4
m2, a = 5
m2 end
~~~

This came up when I read rexxpg how GUARD OFF WHEN is supposed to work and
couldn't make much sense out of what it says: "if you specify "GUARD OFF
WHEN expression," the active method keeps running until expression becomes
true. To become true, another method must assign or drop an object variable
that is named in expression. Whenever an object variable changes, Rexx
reevaluates expression. If expression becomes true, GUARD is turned off,
exclusive use of the variable pool is released, and other methods needing
exclusive use can begin running. If expression becomes false again, GUARD
is turned on and the active method regains exclusive use."

Especially "keeps running" and the last sentence don't seem right to me.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Problem with the ticker test

[Erich Steinböck]

I've fixed the Ticker test group to always cancel the Ticker instance even
when an assert fails.
This should fix the hangs.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Higher precision time on Windows.

[Erich Steinböck]

Hi Rick,
we've so far refrained from using a Windows API younger than 0600 (0600
means no more support for XP, but Vista and above still supported).
GetSystemTimePreciseAsFileTime is _WIN32_WINNT_WIN8 and thus would be the
end of ooRexx support for Vista and Windows 7.

It's the compile define _WIN32_WINNT=0x0600 in CMakeLists.txt that stops us
from inadvertently introducing code with such a newer API.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] A current attempted port to z/OS ...

[Erich Steinböck]

|  common/platform/unix/SysThread.hpp:70:36: error: no matching constructor
for initialization of 'pthread_t'  SysThread() : attached(false),
_threadID(0) { ; }
|  common/platform/unix/SysThread.hpp:86:17: error: cannot convert
'pthread_t' to 'uintptr_t' (aka 'unsigned long') without a conversion
operator
| common/platform/unix/SysThread.hpp:89:37: error: cannot convert
'pthread_t' to 'size_t' (aka 'unsigned long') without a conversion operator
inline size_t hash() { return (((size_t)_threadID) >> 8 * 37; }

I have opened [bug:#1876] and committed some code fixes.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Trademarks

[Erich Steinböck]

Hi René,

persons employed by IBM at the time, and in charge of Object Rexx
>
but then would IBM trademarks not have been for maybe "oRexx or "Object
Rexx", but not "ooRexx" or "Open Object Rexx"?
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

[Oorexx-devel] Trademarks

[Erich Steinböck]

Our docs state "Open Object Rexx™ and ooRexx™ are trademarks of the Rexx
Language Association".

I've searched several online Trademark Portals (e. g. TESS at
https://tmsearch.uspto.gov/bin/gate.exe?f=searchss=4810:74iqlo.1.1)
and cannot verify whether this is actually true.

Is "Open Object Rexx" and/or "ooRexx" (still) a trademark of RexxLA?

Our ooRexx icon, which is set as the rexx.exe icon on Windows also shows
"TM".
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] ooRexx /dev/null for redirected command handler

[Erich Steinböck]

>
> errArr=.array~new
> address system 'some command' with error using (errArr)
>
> With our new twinkle smiley syntax, this of course can be shortened to

address '' 'some command' with error using ((,))
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] OORexx beta - Unable to load library "RXMATH

[Erich Steinböck]

>
> Is case O/S dependent?

The rxmath library is all lowercase on all OS's
but ::requires rxmath library will work on Windows and not on your typical
Unix for well-known reasons.
So for compatibility, ::requires "rxmath" library should be used on all OS's
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] OORexx beta - Unable to load library "RXMATH

[Erich Steinböck]

Hi Cyril,
the library is called rxmath (all lowercase), not RXMATH.
To load it, use ::requires "rxmath" library

(CLASSPATH isn't used with ooRexx)
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Problem with redirected command handler

[Erich Steinböck]

This works as designed. Redirection objects most be accessible at the time
the command is executed.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Add fish to the list of Unix ADDRESS environments

[Erich Steinböck]

Hi Leslie,
I think when this was last discussed it was agreed that nowadays typically
sh would be a link to the preferred shell.
Like on my old Ubuntu it is /bin/sh -> dash

With this ooRexx won't have to create new environments dedicated to a
specific shell and users could consistently code address sh or address ""
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

[Oorexx-devel] checking haveHaltCondition

[Erich Steinböck]

It would be beneficial if long-running tasks like writing/reading/sleeping
implemented in native code could check whether a Halt condition was raised
and return early if so.

How does code like SysFile::write or SysThread::longSleep get access to its
Activation settings, so that it can call haveHaltCondition to check?

Thanks!
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Question ad Write{Output|Error}Buffer

[Erich Steinböck]

>
> how about using WriteOutput()/WriteError() to supply binary data?

WriteOutput and WriteError write the supplied data unmodified to any
redirection object.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Two questions ad AreOutputAndErrorSameTarget()

[Erich Steinböck]

>
> if AreOutputAndErrorSameTarget() is true, does it matter which write API
> one uses
> in this case, e.g. WriteOutput(Buffer) and/or WriteError(Buffer)?
>
You can use either, but shouldn't use both.

In shells it is possible to redirect stdout to stderr ("program
> >stdout_data.txt 2>&1"),
> but also the other way around, hence stderr to stdout ("program
> 2>stderr_data.txt 1>&2").
> Is it possible to distinguish the two from each other?
>
No, ooRexx command redirection is to actual ooRexx objects, there is no
notion of redirecting stdout to stderr (or vice versa).
So these redirection objects are either the same, or not.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Observation with .routine's callwith and threadContext->RaiseException(...)

[Erich Steinböck]

>
> call ppJavaExceptionChain(co,.true)
>
> This look suspicious to me and may be unintended.
Here ppJavaExceptionChain is called with a single argument, an Array of
(co, .true)

I think it's a bit unfortunate that we allow this kind of syntax (as it's
almost always unintended).
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Documentation

[Erich Steinböck]

Fixed with revision [r12464]

On Mon, Jul 4, 2022 at 9:15 AM rvm--- via Oorexx-devel <
oorexx-devel@lists.sourceforge.net> wrote:

> Very minor point:
> orxncurses.pdf preface refers to the Sockets classes.
>
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Java 8 -> 11

[Erich Steinböck]

Hi P.O., we had this discussion back in April and I had found that docs say
"If you’re upgrading your Jenkins controller to run on Java 11, you also
need to upgrade the JVM on your agents".
I don't know if Java 11 is available for all our current agents.


P.O. Jonsson  schrieb am Di., 14. Juni 2022 16:53:

> I intend to have a go at upgrading the Controller (Master) to Java 11
> later this week, this might lead to Jenkins being offline
>
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Ad current samples\ole\apps

[Erich Steinböck]

| silence is taken as approval
Rony, I've already said multiple times and in no unclear words that my
silence just means I haven't written an answer and in no way is an approval.
I'm really getting tired of this.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Problem with address when passing a filename as a command that contains parantheses

[Erich Steinböck]

Another CMD bug.
You'll have to escape the parenthesis with ^
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Jenkins

[Erich Steinböck]

Hi P.O.,
docs say "If you’re upgrading your Jenkins controller to run on Java 11,
you also need to upgrade the JVM on your agents".
I don't know if Java 11 is available for all our current agents, like on
the zLinux.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Trapping SYNTAX on a REPLY thread

[Erich Steinböck]

>
> have base method of the new thread issue it's own signal on syntax
>
If the replied method issues its own signal, it will also have to handle it
itself which isn't useful in the test framework case.

I guess I missed the fact that at the time the replied method runs,
everything else (here: all of the framework) may have already ended,
leaving the method floating in space all alone.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

[Oorexx-devel] Trapping SYNTAX on a REPLY thread

[Erich Steinböck]

When I tried to enhance our test framework to support assert failures from
concurrent threads that were started from test methods, I noticed that a
syntax error in the replied thread of a method cannot be trapped.  Try
running below code with either .c~before or .c~after.

That would mean the concurrent thread has no caller.  Is that how it should
be?
Any ideas how to still trap a syntax in the concurrent thread?

~~~
signal on syntax
.c~before -- or .c~after
return
syntax:
say "captured" condition("object")~code

::class c
::method before class
  say 1/0
  reply

::method after class
  reply
  say 1/0
~~~
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] String method caselessEquals() is documented; which ooRexx implements it?

[Erich Steinböck]

Hi Leslie,
caselessEquals is available as a method of the String class only, not as a
routine as you try to run it
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] my wish: set the option "numeric digits ##" to a single varaible

[Erich Steinböck]

Hi Franz,
if you need higher precision for just one clause, revert precision back to
the default in the next clause, using NUMERIC DIGITS.

To really speed up calculating a list of primes, use the Sieve of
Eratosthenes, which will be faster by several magnitudes.

~~~
-- sieve
use strict arg n = 2015177

prime = .Array~new(n)~fill(.true)~~remove(1)

do i = 2 to sqrt(n)
  if prime[i] = .true then
do j = i * i to n by i
  prime~remove(j)
end
end

primes = prime~allIndexes
say primes~items "primes up to" n
say primes~section(1, 20)~makeString(, " ") "..."
say "..." primes~section(primes~items - 20)~makeString(, " ")

::routine sqrt public external "library rxmath RxCalcSqrt"
~~~
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] ooRexx -> Android / Resuming the porting

[Erich Steinböck]

>
> Would someone be so kind and tell me how to deactivate this test?
>

Hi Thomas,
this may be a CMake issue (CURSES_INCLUDE_PATH is internal to FindCurses).
To remove the test try replacing the include(FindCurses) line in
CMakeLists.txt with these two lines
SET (CURSES_HAVE_NCURSES_H FALSE)
SET (CURSES_HAVE_NCURSES_NCURSES_H FALSE)
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] What am I missing?

[Erich Steinböck]

Hi Gil,
rexxref could be clearer here.

When specified with the TRACE keyword instruction, trace option L traces
internal labels only.
When specified with the ::OPTIONS directive, as a recent enhancement, trace
option L also traces routine and method calls.
So you need to specify ::OPTIONS TRACE L
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Jenkins build machine

[Erich Steinböck]

Hi P.O.,
I've just updated all plug-ins - so we should be clean now.

The only plug-in which has a newer version is "Pipeline: Job", but this
newer version needs Jenkins 2.320, while we're on 2.319.
Why did we go for 2.319?
Jenkins now doesn't seem to show an update notice for Jenkins itself (to
2.320 or newer)
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Question ad changes to ProgramMetaData.{c|h}pp

[Erich Steinböck]

Hi Jon,
correct. No further change should be done as METAVERSION has already been
increased for 5.0.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Question ad changes to ProgramMetaData.{c|h}pp

[Erich Steinböck]

Rony,
METAVERSION was already changed to 42 for 5.0
no double patch please
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] ooRwxx R5 Beta

[Erich Steinböck]

Hi Walter, thanks for spotting this issue.
I checked the build log, but couldn't identify what went wrong.
I've just removed this single installer.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Ad "_DEBUG" for non-Windows builds (Re: RFC: "config" method for .RexxInfo

[Erich Steinböck]

Rony, this should now be fixed with [r12338].
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Ad "_DEBUG" for non-Windows builds (Re: RFC: "config" method for .RexxInfo

[Erich Steinböck]

This NOT the way to fix this.
There is no CMake build type DebWithRelInfo - there is one called
RelWithDebInfo
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] RFC: "config" method for .RexxInfo

[Erich Steinböck]

as already stated in [feature-requests:#796] I don't think that we should
add a RexxInfo debug method

I just committed revision [r12338] to add a _DEBUG define also for
non-Windows builds.
Also note that _DEBUG is non-standard, NDEBUG conforms to the C standard
CMake sets NDEBUG for all except DEBUG builds.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] RFC: "config" method for .RexxInfo

[Erich Steinböck]

I just fixed the missing _DEBUG define for non-Windows builds with revision
[r12334].

On Mon, Nov 29, 2021 at 5:08 PM Rick McGuire  wrote:

> It looks like the _DEBUG_ flag is only getting set for the Windows build
> and not the others. This is likely a CmakeList.txt problem that effects
> more than just rexx -v.
>
> Rick
>
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

[Oorexx-devel] RexxAllocateMemory

[Erich Steinböck]

On Windows RexxAllocateMemory uses GlobalAlloc, whose docs state "The
global functions have greater overhead and provide fewer features than
other memory management functions. New applications should use the heap
functions .."

The mentioned heap functions are HeapAlloc/HeapFree, but require an initial
HeapCreate and a final HeapDestroy.

On Unix RexxAllocateMemory uses standard malloc.

Might it be worth migrating away from GlobalAlloc?  If so, to which of the
two options?
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Reusing an instance of the .Stream class

[Erich Steinböck]

>
> can I reuse fileStream somehow?  There seems to be no way to associate
> fileStream with a different file
>
Hi Leslie,
in fact this can be done by calling INIT directly.

~~~
s = .Stream~new("")
do i = 1 to 99
  s~init(i)
  s~~open("write replace")~~charOut(i)~close
end
~~~

I occasionally considered doing this (e. g. for File or DateTime objects)
and have asked a respective question a few months ago on the Dev list.
I really don't know if calling INIT repeatedly is a supported concept - at
least we don't run any tests for it.
Until we know better, I couldn't recommend using it.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Jenkins

[Erich Steinböck]

Hi P.O,

> If this is an option I can put this machine on a separate switch and make
> it available with short term (1/2 day) notice.
> As it is now it is idle 99% of the time and it is a waste of energy to
> have it up all the time.
>
I totally understand and agree.

If easier to implement you might make the VMs available for e.g. just one
hour per day on midnight.
Any Jenkins jobs would wait until the VM becomes available at the given
time.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Planned commits

[Erich Steinböck]

>
> "silence counts as approval"
>
Rony, my silence is never consent, please stop suggesting that. I just
didn't get around to answering.

concurrency trace
>
I'm in favor of the concurrency trace feature request, but I dislike the
suggested use of an environment variable. I'd rather use a new ::option
directive subkeyword like ::option trace concurrent, or maybe a new trace
prefix, like trace !r.  Better ideas highly welcome.
Also I'd prefer a more compact representation (less white space).  Then I'm
not sure why we'd trace activation numbers or about the "reserve" concept.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Side effects of the recent rollback for Windows?

[Erich Steinböck]

>
> I am not sure how this should be fixed
>
I fixed the test to skip the assert if the WMI interface cannot determine a
default printer

all 6 Windows builds (32 and 64) are reported as failures for this reason
> meaning no fresh Win builds are uploaded to Sourceforge
>
Uploads should still happen.  Build artifact uploads depend on a successful
build, not as successful test
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Question ad native API and resolving an environment symbol

[Erich Steinböck]

Rony,
the FindPackageClass API should do what you want.
It doesn't just resolve a package name, but will resolve any dot variable
name.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Side effects of the recent rollback for Windows?

[Erich Steinböck]

Hi P.O., the VM'ed Windows shows the default printer of the host from which
it is being executed.
This is a situation the test case cannot easily handle, but in a typical
(non VM'ed) scenario it will work and successfully tests
SysWinGetDefaultPrinter.
I've fixed the NOVALUE issue and added a message that the test may fail in
run in a VM.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] ooRexx > Android, Debugging

[Erich Steinböck]

>
> I cannot build 12269 for *Linux*. Until now, I have never had a problem
> with building for Linux. I get this error when linking rxap
>
Hi Thomas, please try building with the latest revision.
Regarding the Linux build issue, you might try creating a fresh build
directory where you re-run CMake and thus build from scratch
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] ooRexx > Android, Debugging

[Erich Steinböck]

>
> This should not be returning an error at all, since this is just a nested
> request for the semaphore.
>
Rick, this should be fine.  The given test runs tests for all three types –
NORMAL, ERRORCHECK, and RECURSIVE – and the ERRORCHECK section is expected
to return EDEADLK/EBUSY upon re-locking.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] ooRexx > Android, Debugging

[Erich Steinböck]

>
> problems getting the configuration for the semaphores to work correctly
>
There really aren't a lot of configuration options any more.  I just
recently cleaned up the SysSempahore code with revision 12259.
And I now believe that we cannot run at all without both
pthread_mutexattr_settype and PTHREAD_MUTEX_RECURSIVE so I might as well
remove that option too (all of our platforms that we currently run builds
on have support for those.)

Thomas, are you at revision 12259 or later?  If not can you please update?

The last time you reported the CMake status of the various pthread
functionality, you showed some "failed"
-- Looking for pthread.h
-- Looking for pthread.h - found
-- Looking for pthread_mutexattr_settype
-- Looking for pthread_mutexattr_settype - not found
-- Looking for pthread_mutex_timedlock
-- Looking for pthread_mutex_timedlock - not found
-- Performing Test HAVE_PTHREAD_MUTEX_ERRORCHECK

*-- Performing Test HAVE_PTHREAD_MUTEX_ERRORCHECK - Failed *-- Performing
Test HAVE_PTHREAD_MUTEX_RECURSIVE
-- Performing Test HAVE_PTHREAD_MUTEX_RECURSIVE - Failed
What's the current status?

To test the actual mutex functionality on Android, can you compile and run
this test?
https://android.googlesource.com/platform/system/extras/+/froyo/tests/bionic/libc/bionic/test_mutex.c
Please also test whether " #define __USE_UNIX98 1" is required to
successfully compile/run.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] [RexxLA] Dir or Not

[Erich Steinböck]

 Hi Andy,


> without the quotes gives the correct response from the function call,
> but that would be an inconsistent use within Linux. I can accommodate
> that, but it seems wrong somehow
>
This is working as designed.  No quotes are required (allowed) as the
folder name really is
Documents and Letters
(without single quotes) which is what Rexx expects.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Errors building mod_rexx.so

[Erich Steinböck]

Hi Terry!

Should RexxVariablePool be linked into my .so file?  Or should it be loaded
> at runtime?
>
Loaded at runtime, I'd say

/usr/bin/ld: cannot find -l/usr/local/lib/librexx.so
> /usr/bin/ld: cannot find -l/usr/local/lib/librexxapi.so
>
I don't think -l accepts a path or an extension.  If required, you could
use -L /usr/local/lib -lrexx -lrexxapi

/usr/lib/apache2/modules/mod_rexx.so: undefined symbol: RexxVariablePool
>
This link might be relevant
https://stackoverflow.com/questions/22342348/how-to-load-dependencies-external-libraries-for-apache2-modules
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Jenkins: Rockhopper upgraded to CMake 3.20.3 and added local Subversion version

[Erich Steinböck]

Hi René, success .. this works now.
Thanks!!!
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Jenkins: Rockhopper upgraded to CMake 3.20.3 and added local Subversion version

[Erich Steinböck]

I don't know if this works with options specified together with the allowed
command
Are you sure you did sudo visudo -f /etc/sudoers.d/rvjansen as there seems
to be no  /etc/sudoers.d/rvjansen file

ls -l /etc/sudoers.d

-rw-r--r-- 1 root root  31 Sep  5  2020 devops
-r--r- 1 root root 958 Jan 18  2018 README
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] ooRexx > Android

[Erich Steinböck]

>
> I dont have access to /tmp on the device
>
Can you check if the TMPDIR environment variable is defined (echo $TMPDIR)
and if so run rxapi like this
XDG_RUNTIME_DIR=$TMPDIR rxapi
Can you check if /data/local/tmp is available and files can be created in
it and if so run rxapi like this
XDG_RUNTIME_DIR=/data/local/tmp rxapi
Else try to find any directory that allows creating files and run rxapi as
described above

but Im working on a hint Prof. Flatscher gave me hoping to get around this
> issue
>
Will you share it with us?
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Jenkins: Rockhopper upgraded to CMake 3.20.3 and added local Subversion version

[Erich Steinböck]

Nope, sudo doesn't work - neither for dpkg --install nor for --remove
It fails with "sudo: no tty present and no askpass program specified"

When you try it from the command line, does it work for you?
Is it now set to " rvjansen ALL = (root) NOPASSWD: /usr/bin/dpkg" ?
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Jenkins: Rockhopper upgraded to CMake 3.20.3 and added local Subversion version

[Erich Steinböck]

Hi René,
now sudo seems to work for dpkg --remove, but doesn't for dpkg --install
Did you explicitly restrict sudo to dpkg --remove ?
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Jenkins: Rockhopper upgraded to CMake 3.20.3 and added local Subversion version

[Erich Steinböck]

>
> sudo dpkg --remove oorexx
> sudo: no tty present and no askpass program specified
>
> Hi René, it doesn't work yet.

Did you do:
sudo visudo -f /etc/sudoers.d/rvjansen
rvjansen ALL = (root) NOPASSWD: /usr/bin/dpkg
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Jenkins: Rockhopper upgraded to CMake 3.20.3 and added local Subversion version

[Erich Steinböck]

Hi René, thanks!

Our build jobs now also test installation of the built binary package.
On Ubuntu this means running dpkg --install and dpkg --remove
But these require sudo to work.  Would you mind giving the rvjansen user
sudo dpkg without password?

Thanks and regards, Erich
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] ooRexx > Android

[Erich Steinböck]

Hi Thomas!

executing rexximage on my Linux works fine and builds rexx.img
>
OK, please revert the changes you did to Activity::resolveProgramName

root   14888  14843 10758192  3336 0 0 R ps
>
What's the reason you are building with root instead of a standard user?

rxapi: lockfile is locked by another rxapi instance; exiting

Does /tmp/.ooRexx-5.0.0-64-root.lock exist?
Can you rm /tmp/.ooRexx-5.0.0-64-root.lock and try again?
Can you try running rxapi from a different (root or non-root) user?
Might /tmp on the emulated Android device be a file system that is shared
with the host Linux (and be locked there)?
As far as I understand direct access to the file system is typically pretty
much restricted on Android. Can your test users access /tmp?
If all else fails, you may try to write a test program that does the
equivalent of acquireLock in APIService.cpp and try whether locking files
on various paths succeeds
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] ooRexx > Android

[Erich Steinböck]

> rxapi is found. Nevertheless the program hangs
Are we taking about running rxapi ON ADNROID?

Can you kill any running rxapi tasks and then manually run rxapi (just type
./rxapi)
What output do you see?  Something along these lines is expected:
rxapi: lockfile path is /tmp/.ooRexx-5.0.0-64-user.lock
rxapi: lockfile lock acquired
rxapi: service path is /tmp/.ooRexx-5.0.0-64- user .service
rxapi: service successfully started; listening
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] ooRexx > Android

[Erich Steinböck]

Hi Thomas,
I thought in your previous mail you said that running ./rexximage rexx.img
*ON ANDROID* fails with "file is unreadable"
This is why I suggested to add some debug printf's.

None of your screenshots in this mail shows the "file is unreadable" error.
Did I misunderstand?
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] [devel] Ad machine/processor architecture information

[Erich Steinböck]

>
> machine/processor architecture
> identify the 'endianness' of the processor
>

I'm in favour of both suggestions, but don't put your hopes too high for
the machine/processor bit.
Basically this will be the output of uname -p with names decided by the
distribution, e. g.
x86_64 (Linux et al.)
amd64 (e. g. OpenBSD, FreeBsd)
i86pc (OpenIndiana)
s390x (ibm)
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Enhanced 'json.cls' suggested as a plug-in replacement for the current 'json.cls' in ooRexx 5.0 beta

[Erich Steinböck]

>
> suggesting to use this 'json.cls' version to replace the old version, if
> there are no objections

I don't like it, not least because it's 50% slower than what we have.

Also please no rip-and-replace code rewrites; if something is broken (.nil
<--> "null" ?) open a bug, or if something needs to get enhanced, open a
feature request.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] ooRexx > Android

[Erich Steinböck]

Hi Thomas!


> Error 3.1:  Failure during initialization: File "" is unreadable.
> Logic error: Error building kernel image.  Image not saved.


It seems as if image build wouldn't find (resolve) the base .orx class
files.
Can you in Activity.cpp replace the code body of
Activity::resolveProgramName with
{
printf("Activity::resolveProgramName name=%s\n", name->getStringData());
RexxString *res = instance->resolveProgramName(name, dir, ext, type);
printf("Activity::resolveProgramName res =%s\n", res->getStringData());
return res;
}

When building the image again, this should print something like below
$ ./rexximage r.img
Activity::resolveProgramName name=CoreClasses.orx
Activity::resolveProgramName res
=/home/user/oorexxbuild/release/bin/CoreClasses.orx
Activity::resolveProgramName name=StreamClasses.orx
Activity::resolveProgramName res
=/home/user/oorexxbuild/release/bin/StreamClasses.orx
Activity::resolveProgramName name=PlatformObjects.orx
Activity::resolveProgramName res =/home/ user
/oorexxbuild/release/bin/PlatformObjects.orx
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Troubles with pthread and Android

[Erich Steinböck]

Hi Thomas,

I excluded the line*es = crypt(str, salt);*
> and I also excluded the whole *SysWordExp Method*.
>
Commenting them is fine. Not all platforms provide full functionality.
These aren't core functions, so they are just left out.
CMake should detect this and skip them if unavailable.
In fact, CMake does detect wordexp, but it seems it doesn't work.
On  a typical system you should see

-- Looking for wordexp
-- Looking for wordexp - found
-- Looking for wordexp.h
-- Looking for wordexp.h - found

What do you see?

I can then update our CMake accordingly.

/bin/sh: line 1:
> /home/tjk/coding/oorexx/oorexxsvn/main/trunk/bin/rexximage: No such file or
> directory
>
rexximage is an executable that creates rexx.img.
What does ls -l /home/tjk/coding/oorexx/oorexxsvn/main/trunk/bin/rexximage
show?
For a test, you can try to run it manually with ./rexximage rexx.img
When doing a cross-compile this executable is probably an Android ARM
binary that cannot be directly run on your host system.
Do you have an Android emulator to run it in?
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Troubles with pthread and Android

[Erich Steinböck]

Hi Thomas!

> Removing ${ORX_SYSLIB_PTTHREAD} in CMakeLists.txt helps a bit but instead
> of getting the error:
> *"cannot find -lpthread"*
>
As far as I can see, Android doesn't have libpthread, so your fix seems
reasonable

"error: use of undeclared identifier 'posix_spawn_file_actions_init'; did
> you mean 'posix_spawn_file_actions_t'?"
>
In spawn.h, where your compiler complains that there is a
posix_spawn_file_actions_t, but no posix_spawn_file_actions_init, I can see
that
posix_spawn_file_actions_init is actually declared in line 81 (see
https://android.googlesource.com/platform/bionic/+/master/libc/include/spawn.h
)
Can you check your actual spawn.h whether it declares it?
Regarding the "__INTRODUCED_IN(28)" declaration: maybe your compile is
restricted to a certain API level below 28
Does "android-ndk-r22b" mean your development toolkit is level 22?

if (ANDROID)
> add_definitions(-DHAVE_PTHREAD_MUTEXATTR_SETTYPE)
> add_definitions(-DHAVE_PTHREAD_MUTEX_RECURSIVE)
> endif ()
>
I don't think this makes sense.  When CMake runs, it checks for available
features (you can see found features in the CMake output) and then sets
those defines appropriately.
A typical run will show below CMake output.  What do you see?
-- Looking for pthread.h
-- Looking for pthread.h - found
-- Looking for pthread_mutexattr_settype
-- Looking for pthread_mutexattr_settype - found
-- Looking for pthread_mutex_timedlock
-- Looking for pthread_mutex_timedlock - found
-- Performing Test HAVE_PTHREAD_MUTEX_ERRORCHECK
-- Performing Test HAVE_PTHREAD_MUTEX_ERRORCHECK - Success
-- Performing Test HAVE_PTHREAD_MUTEX_RECURSIVE
-- Performing Test HAVE_PTHREAD_MUTEX_RECURSIVE - Success


What's the general setup here .. how did you set up your Android tool chain?
Which compiler is this? CLANG?
Is this a cross-compile - host Mac, target Android?

Regards,
E
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Install errors

[Erich Steinböck]

>
> dpkg (subprocess): unable to execute old oorexx package pre-removal script
> (/var/lib/dpkg/info/oorexx.prerm): No such file or directory
> oorexx.prerm does exist and looks OK
>

Hi Bob, I've seen cases like this: oorexx.prerm couldn't run because it has
Windows-style cr-lf line endings.
Either echo "exit 0" > /var/lib/dpkg/info/oorexx.prerm or rm
/var/lib/dpkg/info/oorexx.prerm should fix that - not sure which of those
two worked.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Introduction new List member - OORexx for Android

[Erich Steinböck]

Hi Thomas, one of your Professor's students had at least partially done an
Android port a few years ago. It may be worth checking what (if any)
information is still available from back then.

Will be a step learning curve.

Be assured you can ask any question here .. we may or may not be able to
help. Personally I've got zero experience with Android.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Problem? with OORexx API function ObjectToDouble

[Erich Steinböck]

>
> it should not be difficult to directly generate a double value


You'd think so, but read these stunning strtod() algorithm descriptions (a
must-read for anyone who likes math).
https://www.exploringbinary.com/how-strtod-works-and-sometimes-doesnt/
https://www.exploringbinary.com/how-glibc-strtod-works/
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Problem? with OORexx API function ObjectToDouble

[Erich Steinböck]

Hi Ruurd, instead of defining your latitude/longitude arguments as CSTRING
and then manually converting them, you can define your arguments as (e. g.)
float and have Rexx do the conversion for you.

~~~
float lat, // Latitude in decimal degrees
float lon, // Longitude in decimal degrees
~~~

The actual issue may be a bug as ObjectToDouble calls C's strtod() which
seems to adhere to the locale (decimal dot vs decimal comma).
It seems rexx doesn't globally set the "C" locale but runs with whatever
locale it is being started in.

Does it work as expected if you run your script in either of these two
forms?
LANG=C rexx 
LC_NUMERIC=C rexx 
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

[Oorexx-devel] Converting size_t, ssize_t

[Erich Steinböck]

I think our API doesn't provide conversion functions for the types size_t
and ssize_t, so we have to use either (Unsigned)Int32 or (Unsigned)Int64 to
convert.

I still want to use size_t and ssize_t and would have expected this to
work.

64-bit conversions and 32-bit size_t conversion works as expected, but on
Windows 32 bit the ssize_t conversion compiles with error C2664: 'logical_t
RexxCallContext_::Int32(RexxObjectPtr,int32_t *)': cannot convert argument
2 from 'ssize_t *' to 'int32_t *'

~~~
RexxObjectPtr o;
size_ts;
ssize_t   ss;
context->UnsignedInt32(o, );
context->Int32(o, );
~~~

Why is that?
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] mod_rexx Linker Unresolved Symbols...

[Erich Steinböck]

Hi Terry,
you can find Dave's mod_oorexx makefiles at
https://sourceforge.net/p/oorexx/code-0/HEAD/tree/incubator/mod_oorexx
Maybe you can work from there.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

[Oorexx-devel] IP V6 support

[Erich Steinböck]

What was the reason why [r11452]
 removed the IP V6 classes
socket6, smtp6 etc. from the 5.0 trunk?  Not yet ready-to-ship?
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Errors compiling mod_rexx

[Erich Steinböck]

Hi Terry,

I'm having more problems building mod_rexx
>
I checked the mod_rexx source on SF.  You can look up ooRexx exit function
protoypes and examples in rexxapi.pdf at
https://sourceforge.net/projects/oorexx/files/oorexx-docs/5.0.0beta/

E. g. 1.15.1. "Writing Context Exit Handlers":

~~~
int REXXENTRY Rexx_IO_exit(
RexxExitContext *context, // the exit context API vector
int exitNumber, // code defining the exit function
int subfunction, // code defining the exit subfunction
PEXIT parmBlock); // function-dependent control block
~~~

Or in 2.7.3.1.3. "Example":

~~~
int SetRexxVariable(
const char *name, /* Rexx variable to set */
char *value) /* value to assign */
~~~

is there some standard (portable) type that I should be using instead of
> LONG?
> is there some standard type I should use in place of PSZ?
>
Above suggests using char * instead of PSZ, and int instead of LONG.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Statusreport new testGroups

[Erich Steinböck]

Hi P.O.,

> Since the Jenkins Master machine is in my corner I could give it a try,
> I use Virtualbox and have built on various Linux, would Virtualbox be an
> acceptable solution?
>
I know nothing about Virtualbox but I think we have only very modest
demands on the VM capabilities.
Any common RPM-based Linux would be great. We've just had an issue reported
against our RPM installers and can't test it.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Statusreport new testGroups

[Erich Steinböck]

> I noticed you made it much more advanced - do you still want me to amend
> the rest of the test cases or do you prefer to go over them yourself?
>
Yes, please you do it.

Would one test group „samples.testGroup“ be ok, or do you want them split
> in any way?
>
It's fine either way. You might want to have a few groups, like the ones
which require a supporting .cls, or some which are very different (in
whatever respect) than others.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Statusreport new testGroups

[Erich Steinböck]

>
> I wonder if the last line *options all syntax* should be in or not?
> Anything else?
> Regarding cd/rexx: for this particular test case you need to be running it
> from where the samples actually reside
> gets stuck quite frequently on Windows (worse for 32 bit than 64 bit) and
> the same for Debian
> I will deal with the other test cases when you give this one green light.
>

Hi P.O.,
I think I've fixed the hang issue
Also, I've changed the sample such that it can be run from any directory
::options all syntax raises syntax errors for conditions that else go
unnoticed, which sometimes helps finding error.
I've also updated the test Group - everything should be fine now

When updating your existing test groups for individual samples you may
consider grouping sample tests together into a single testGroup to reduce
maintenance efforts
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Building ooRexx on Windows

[Erich Steinböck]

>
> Hi P.O. !


> Can someone point me to the correct place (or places)?
>
It should be put into the directory that vcvarsall.bat adds to PATH. For a
64-bit build this would typically be ..\Hostx64\x64 and ..\HostX86\x86 for
a 32-bit build. To be sure, just check PATH after calling vcvarsall.
Last time I installed it was sufficient to copy just jom.exe.  If this has
changed since, please let me know what else is required, so I can update
the Wiki accordingly.

the link to VS 2017 is dead it seems, I used VS 2019 instead from
> https://visualstudio.microsoft.com/free-developer-offers/
> maybe someone can amend the WIKI?
>
I updated our Wiki.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Two questions ad samples\ole\apps

[Erich Steinböck]

>
> maybe its two preceding lines should not be shown
>
Hi Rony, as far as I can see the trace correctly shows where the issue
arises.
This is what I'd expect.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Suggested samples\ole\apps\samp09.rex (Re: Two questions ad samples\ole\apps

[Erich Steinböck]

Hi Rony, just go ahead.
Maybe rather call the created Excel samp09 instead of demo - less chance of
a clash and relates better to samp09.rex
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

[Oorexx-devel] NewString timings

[Erich Steinböck]

While working on the Stream performance issues, I noticed we have a
performance hit on Windows for calls to the NewString API vs. what we had
in 4.2. (Stream is implemented through the APIs, so NewString has quite an
impact).

By building past revisions I traced it down to two commits, r11494 and
r11503, which are both adding a significant amount of cycles to this API
call, bringing total clock time up from some 1.5 usec to some 4.2 usec per
call.

See attached chart. Timings were done with Windows high-resolution timers,
which occasionally show random spikes, but the overall picture seems to be
consistent. Timings were done for both routine and method contexts.

Both r11494 and r11503 are rather large commits, touching the dispatcher
and other areas and it's not obvious for me what may be the cause.

Are there any ideas what parts of these commits are introducing the
increase?


I didn't run the same tests on Linux, but I think this may be a
Windows-only issue as NewString timings seem to be faster and unaffected by
above commits, though as I said, I only have circumstantial evidence for
this.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

[Oorexx-devel] Rexx LA Mailing List

[Erich Steinböck]

ooRexx docs say: "The Rexx Language Association maintains a forum at
http://www.rexxla.org/forum.html with its archive located at
http://rexx-language-association-forum.44760.x6.nabble.com.;

Currently the two links don't seem to show anything useful.

Is this an issue we can fix, or should this sentence be removed from the
docs?
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Statusreport new testGroups

[Erich Steinböck]

> if we set up a VM to run on the Jenkins master itself, there is plenty of
capacity left
I'd love to see Jenkins covering additional distributions.
Is there someone who wants to (can) set up Fedora, OpenIndiana, OpenBSD,
OpenSUSE Leap whatever other distribution we don't yet run regular tests on?
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Statusreport new testGroups

[Erich Steinböck]

 > Regarding your more recent change to the framework: It seem these
changes did not give the same functionality on Linux, several samples test
cases are now broken
Right, here's is why:
The changes I've done should enable us to run tests directly off the build
tree without having to install (and without touching any installed ooRexx).
This will help us run on slaves even when we don't have sudo (we really
shouldn't require no sudo any more).
To test whether things work as envisioned I've actually removed the
installed ooRexx from my Ubuntu VM.
This is why I believe running the samples through execRexxPrg() and not via
cd / rexx will fix this.

I guess I'll need to verify this.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Statusreport new testGroups

[Erich Steinböck]

Hi P.O., re your recently committed additional test groups for Samples:

all testGroup files should have set these three SVN properties:
 svn propset svn:eol-style native name.testGroup
 svn propset svn:executable on name.testGroup
 svn propset svn:keywords "Date Rev" name.testGroup

Also testGroup files are expected to have proper headers - exactly like
below:
#!/usr/bin/env rexx
/*
  SVN Revision: $Rev$
  Change Date:  $Date$
*/

Your testGroup template defines global variables like shouldStop that
aren't actually used anywhere.
This includes a variable stopStop which is actually a typo.

There's no need to special-case ADDRESS "" vs. ADDRESS COMMAND.
Cross-platform ADDRESS formats include ADDRESS SYSTEM or ADDRESS ""

The main tests run the samples with an explicit REXX external command. They
should be run with execRexxPrg() instead.

Your test groups typically run the tested sample twice, which shouldn't be
necessary. Once run, return code and output tests can be done at the same
time. There's also no need to test for a  return code (this was just an
internal quirk that I've removed).

Also to simplify maintenance you might check out the template that I just
committed: wmi.testGroup uses a single test per sample, making the tests
easier to understand and more concise.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

[Oorexx-devel] INIT

[Erich Steinböck]

While rexxref warns that UNINT "should not be invoked directly", I don't
see similar warnings regarding INIT usage.

We have classes like File or DateTime which are almost unmutable: there's
no obvious method to change the instance - just new instances can be
created.  Except there's INIT.  By using INIT directly, the instance *can*
be changed. I've had cases where this INIT trick would have come really
handy.

Do we regard calling INIT directly as supported, or a no-no?
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] No inherit for REXX package classes

[Erich Steinböck]

> It really is a security feature.
> Things start getting terribly wonky if the behavior of a class gets
altered beyond the author's intentions.

What about allowing just something most basic, like *adding *new methods
only (no deletions or modifications of existing)?
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

[Oorexx-devel] No inherit for REXX package classes

[Erich Steinböck]

For an elegant solution I so often wish I could add a method to a class
within the REXX package, but can't, as it's forbidden.  I think it's also
impossible to add a method to an existing instance of such a class (an
instance like e. g. a DateTime that is returned from a .File method).

What is preventing us from allowing this?
Is this a technical thing (because these classes are in the image)?
Or is it meant to be a security feature?

Any ideas how we could allow this (or similar)?
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Re: [Oorexx-devel] Fwd: Segfault

[Erich Steinböck]

>
> how to get a core dump out for further analysis in Linux after a
> segmentation fault?
>

I'm no authority in debugging segfaults, but I think it was agreed that
core dumps are of little use.
To produce output that may be more useful follow the steps documented in
https://sourceforge.net/p/oorexx/wiki/how-to-debug-oorexx/#debugging-with-gdb

Segmentation faults during testOOrexx runs have been especially difficult
to track down when related to garbage collection issues.
___
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel