Re: [Oorexx-devel] I/O again

2009-04-27 Thread Mike Cowlishaw
Sorry to report that the dreaded 'r' still appears with Rick's latest (build 4507). sigh Maybe best to just document this in the readme as a probable Windows problem? Everything else is working beautifully! Mike - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - Mike

Re: [Oorexx-devel] I/O again

2009-04-27 Thread Rick McGuire
Ok, I have one more thing to try. I'm starting to think this might be a memory-management issue, not an I/O issue. The buffer used for the I/O is one of the few places in the ooRexx code where malloc() is used to allocate memory. In fact, it wouldn't surprise me if this is the first malloc

Re: [Oorexx-devel] I/O again

2009-04-27 Thread Mark Miesfeld
I've applied the patch and am building the install package right now. I'll let you know when it's ready Mike. Rick, I'm hopeful about this. On Windows I really think we should only use GlobalAlloc and its relatives instead of malloc. The only time I use malloc is when I'm going to free it right

Re: [Oorexx-devel] I/O again

2009-04-27 Thread Mark Miesfeld
Okay, it's built. I sent Mike the link to get it. -- Mark Miesfeld On Mon, Apr 27, 2009 at 5:48 AM, Mark Miesfeld miesf...@gmail.com wrote: I've applied the patch and am building the install package right now. I'll let you know when it's ready Mike. Rick, I'm hopeful about this.  On Windows

Re: [Oorexx-devel] I/O again

2009-04-27 Thread Rony G. Flatscher
Pretty regularly I can get that problem too when opening a command line window, and first thing I would do is using rexxtry and immediately issuing a say statement (typing while rexxtry.rex starts up). Here's an example that works *quite reliably* for me: Microsoft Windows XP [Version

Re: [Oorexx-devel] I/O again

2009-04-27 Thread Rick McGuire
Have you tried reproducing using the patched version? I tried this technique this morning both the unpatched and official beta version and didn't get any failures. Rick On Mon, Apr 27, 2009 at 10:42 AM, Mark Miesfeld miesf...@gmail.com wrote: Well, okay.  By pasting lines into a console

Re: [Oorexx-devel] I/O again

2009-04-27 Thread Mark Miesfeld
I'm just about to try that. I was seeing how reliable it was that I saw RXIT. Seems to be ever time when I paste the text into the window. I think another key is that Rony says he sees it when: immediately issuing a say statement (typing while rexxtry.rex starts up). So here, the keyboard

Re: [Oorexx-devel] I/O again

2009-04-27 Thread Rick McGuire
Hmmm, this is starting to push the issue back to being an I/O problem. Here's something to try. In SysFile::setStdIn(), try commenting out the line _setMode(fileHandle, _O_BINARY); and see if you can recreate the problem. I'm wondering if switching the mode while characters are already in

Re: [Oorexx-devel] I/O again

2009-04-27 Thread Mark Miesfeld
On Mon, Apr 27, 2009 at 7:47 AM, Rick McGuire object.r...@gmail.com wrote: Have you tried reproducing using the patched version?  I tried this technique this morning both the unpatched and official beta version and didn't get any failures. Yes, I just tried with the patched build I did

Re: [Oorexx-devel] I/O again

2009-04-27 Thread Mark Miesfeld
Okay, I'll try that. Tried with 3.2.0 about 10 times and it never happened. On 4.0.0, it happens for me every single time I paste the lines into the console window. -- Mark Miesfeld On Mon, Apr 27, 2009 at 8:01 AM, Rick McGuire object.r...@gmail.com wrote: Hmmm, this is starting to push the

Re: [Oorexx-devel] I/O again

2009-04-27 Thread Rick McGuire
Ok, I've got the magic incantation correct now too. The key was not picking up the EOL for the exit command but physically hitting enter. I seem able to reproduce this reliably now too. Rick On Mon, Apr 27, 2009 at 11:05 AM, Mark Miesfeld miesf...@gmail.com wrote: Okay, I'll try that. Tried

Re: [Oorexx-devel] I/O again

2009-04-27 Thread Mark Miesfeld
On Mon, Apr 27, 2009 at 8:09 AM, Rick McGuire object.r...@gmail.com wrote: Ok, I've got the magic incantation correct now too.  The key was not picking up the EOL for the exit command but physically hitting enter. I seem able to reproduce this reliably now too. That's good. I tried

Re: [Oorexx-devel] I/O again

2009-04-27 Thread Rick McGuire
This is useful. I now have a consistent means of reproducing this, I'm able to gather a little more debug info. Unfortunately, what I've gotten so far just confirms what I already new...that the bogus character is coming from the C runtime routine. RIck On Mon, Apr 27, 2009 at 11:17 AM, Mark

Re: [Oorexx-devel] I/O again .. using Rexxtry

2009-04-27 Thread Mike Cowlishaw
I hadn't thought of trying Rexxtry. Thanks Rony! An even simpler scenario: start Rexxtry, then just type a single letter command (and Enter) that is valid and does nothing (so returns very fast) (I have a c.rex that contains only a comment), and repeat, fast. Below is what I just got ---

Re: [Oorexx-devel] I/O again

2009-04-27 Thread Rick McGuire
I think I've fixed this! A big thanks to Rony...his method for reproducing this was 100% effective once Mark gave me the key to getting it to work. The problem only seemed to occur if our buffering is turned off for stdin. When buffering was turned off, I would consistently see the bogus

Re: [Oorexx-devel] I/O again

2009-04-27 Thread Mark Miesfeld
I saw your commit and am building from trunk right now. (Guess I should have built from the beta.) But, I think trunk is good enough for Mike to just try. -- Mark Miesfeld On Mon, Apr 27, 2009 at 9:20 AM, Rick McGuire object.r...@gmail.com wrote: I think I've fixed this!  A big thanks to

Re: [Oorexx-devel] I/O again

2009-04-27 Thread Mark Miesfeld
It's fixed for me also. The new test build is almost ready, I'll let Mike know as soon as it is available. -- Mark Miesfeld On Mon, Apr 27, 2009 at 9:20 AM, Rick McGuire object.r...@gmail.com wrote: I think I've fixed this!  A big thanks to Rony...his method for reproducing this was 100%

[Oorexx-devel] I'll refresh the beta builds today

2009-04-27 Thread Mark Miesfeld
All, I'm planning on refreshing the beta builds on SourceForge today. All the bugs reported so far have been verified as fixed, except for Mike's of course. I'm hopeful that one is also. It seems there is a good chance of it. I know David is busy, so I'm planning on doing the Linux builds

Re: [Oorexx-devel] Build 4507 - 4521

2009-04-27 Thread Mike Cowlishaw
Something else changed between these builds ... now the first time GoServe calls Rexx it works fine, but on the second call it crashes with a C05 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - Mike Cowlishaw, IBM Fellow http://bit.ly/mfc IBM UK (MP8), PO Box

Re: [Oorexx-devel] Build 4507 - 4521

2009-04-27 Thread Rick McGuire
This could conceivably be caused by the same change, but I'm probably going to need something to recreate this so I can debug. Rick On Mon, Apr 27, 2009 at 12:55 PM, Mike Cowlishaw m...@uk.ibm.com wrote: Something else changed between these builds ... now the first time GoServe calls Rexx it

Re: [Oorexx-devel] Build 4507 - 4521

2009-04-27 Thread Mark Miesfeld
On Mon, Apr 27, 2009 at 9:55 AM, Mike Cowlishaw m...@uk.ibm.com wrote: Something else changed between these builds ... now the first time GoServe calls Rexx it works fine, but on the second call it crashes with a C05 Mike, I did the build from trunk, rather than the beta source tree. Let

Re: [Oorexx-devel] Build 4507 - 4521

2009-04-27 Thread Mark Miesfeld
Rick, I think Mike and I might have got crossed up by slow e-mails. (Maybe.) His subject title says 4521. But, the build with your fix in it was 4523. Anyhow, I pointed him to a 4523 build that is a fresh check out from the beta branch, rather than a build from my trunk. That way we can be

Re: [Oorexx-devel] Build 4507 - 4521

2009-04-27 Thread Mike Cowlishaw
OK, here's update ... I think Mike and I might have got crossed up by slow e-mails. (Maybe.) His subject title says 4521. But, the build with your fix in it was 4523. Anyhow, I pointed him to a 4523 build that is a fresh check out from the beta branch, rather than a build from my trunk.

Re: [Oorexx-devel] Build 4507 - 4521

2009-04-27 Thread Rick McGuire
The only other difference between 4507 and 4523 is the fix for SysFileSearch(). Nothing at all related to rxapi. Rick On Mon, Apr 27, 2009 at 2:32 PM, Mike Cowlishaw m...@uk.ibm.com wrote: OK, here's update ... I think Mike and I might have got crossed up by slow e-mails. (Maybe.)  His

Re: [Oorexx-devel] Build 4507 - 4521

2009-04-27 Thread Mike Cowlishaw
The only other difference between 4507 and 4523 is the fix for SysFileSearch(). Nothing at all related to rxapi. Sorry ... I meant Rexx APIs, not rxapi. I have exactly one call to SysFileSearch in all my Rexx programs, and definitely nowhere near the failing case (in fact I don't even

Re: [Oorexx-devel] Build 4507 - 4521

2009-04-27 Thread Rick McGuire
Most of the commits in that range were to the branch where I tried replacing the the C runtime functions with native ones. Add to that the fact we're updating both the trunk and beta branch with all bug fixes, the commit numbers ratchet up pretty quickly. However, it looks like the version that

[Oorexx-devel] ooRexx 4.0 Beta Refreshed (r4525), Now Available

2009-04-27 Thread Mark Miesfeld
The ooRexx 4.0 Beta files on SourceForge have been refreshed. The subversion level is 4525. All bugs reported during this beta, to date, have been fixed in this build. The files are available at: https://sourceforge.net/project/showfiles.php?group_id=119701 Those people participating in the