INI file corruption after running Java app (JavaRDP) for extended period

[Lewis Rosenthal]

Likely due to memory fragmentation. I know that the condition is triggered by running JavaRDP for extended periods along with my other mix of apps.

The symptom is that after JavaRDP has been running for several hours, even if SeaMonkey has been shutdown and restarted a few times, shutting down gets me all the way to the point where the desktop closes, and then I get a message that changes could not be saved to the INI files, and a prompt (by XWorkplace) to restart the desktop and use the standard shutdown routine. This never works, as invariably, the files cannot be written.

A reboot (even when not saving the files) generally results in either irreparable desktop damage (broken folder positions, file handles, etc.) or the system cannot even locate the desktop directory (even with SET DESKTOP set in CONFIG.SYS and valid EAs on the desktop directory). In these cases, a restore from an archive is the only recovery.

Please let me know what I can provide (some utility to track the fragmentation? Theseus output?) or perhaps what memory setting I might try when starting the JVM to avoid the situation. This seems to have started around GA4, but possibly earlier.

VIRTUALADDRESSLIMIT is set at 1576, but I have tried higher values with no discernable change.

[dmik]

Thanks for the report! The request is the same as in #65: please try to find an OpenJDK release that works for you as this may help us in finding the reason.

[dmik]

Also a question: do you use the Java or the Flash plugin in SeaMonkey when it happens?

[Lewis Rosenthal]

It's hard to tell what's running in SeaMonkey, though it's entirely possible that Flash is loaded for something. SeaMonkey is started from a batch wrapper:

@ECHO OFF

REM *** Some common stuff
SET MOZILLA_DRIVE=J:
SET MOZILLA_BIN=\mozilla.org\seamonkey
SET MOZILLA_HOME=J:\OS2WEB
SET MOZ_PLUGIN_PATH=%MOZILLA_DRIVE%\MOZILLA.ORG\MOZ_PLUG
REM SET MOZILLA_USE_EXTENDED_FT2LIB=T
REM SET NSPR_OS2_NO_HIRES_TIMER=1
SET MOZ_ACCELERATED=0

REM *** Set TZ to something more expected on OS/2
SET TZ=EST5EDT,3,2,0,7200,11,1,0,7200,3600

REM *** PROCDUMP settings ***
REM CALL=PROCDUMP.EXE SET /PROC:seamonkey.EXE /C:0 /P:0 /K:N /D:SUMM,PRIVATE,SHARED,INSTANCE,SYSFS,SYSSEM,SYSIO

REM *** Some Java variables; unrem java_plugin_trace to enable debugging
REM SET JAVA_HIGH_MEMORY=0
REM SET GCD_JAVA_HOME=C:\java141\jre
REM SET JAVA_HOME=C:\Java142\bin
REM SET JAVA_HOME=C:\JAVA141\jre
REM SET PATH=%JAVA_HOME%;%PATH%
REM SET CLASSPATH=
REM SET CLASSPATH=C:\IBMGSK\CLASSES\SRIP.JAR;C:\IBMGSK\CLASSES\SGUIDE.ZIP;C:\IBMGSK\CLASSES\CSSGKEY.JAR;.\.;C:\OS2\JAVAAPPS\LVMGUI.ZIP;C:\TCPIP\JAVA;
REM SET JAVA_PLUGIN_TRACE=1
SET ICEDTEA_WEB_JRE=C:\JAVA160
SET ICEDTEA_WEB_DATA=%MOZ_PLUGIN_PATH%


REM *** mplayer location & path addition
SET MPLAYER_PATH=C:\MMOS2\mplayer\mplayer.exe
SET PATH=C:\MMOS2\mplayer;%PATH%
SET ENDLIBPATH=C:\MMOS2\mplayer;

REM *** Run seamonkey in its own space to allow running FireFox and Nvu separately
REM SET MOZ_NO_REMOTE=1
REM SET LIBPATHSTRICT=T
SET PATH=%MOZILLA_DRIVE%%MOZILLA_BIN%;%PATH%
SET BEGINLIBPATH=%MOZILLA_DRIVE%%MOZILLA_BIN%

REM *** Delete old, possibly incorrect files from profile directories
REM c:\tools\4os2\4os2.exe /c del %MOZILLA_HOME%\mozilla\seamonkey\profiles\pluginreg.dat /q /s /z > nul
REM c:\tools\4os2\4os2.exe /c del %MOZILLA_HOME%\mozilla\seamonkey\profiles\startupCache.4.little /q /s /z > nul

REM *** Some FreeType2 tweaks
SET USE_STYLE=ECS
CALL c:\bats\infinality-settings.cmd

REM *** Now, fire up the lizard
%MOZILLA_DRIVE%
CD %MOZILLA_BIN%
IF "%JAVA_PLUGIN_TRACE%"=="1" GOTO DEBUG

seamonkey.exe %1 %2 %3 %4 %5 %6 %7 %8 %9
GOTO END

:DEBUG
seamonkey.exe %1 %2 %3 %4 %5 %6 %7 %8 %9 >plugin_std.log 2>plugin_err.log
GOTO END

:END

(Some of the above is quite old and should probably be removed or better commented, but you can see what's active and what's not.)

I'll also see about getting a few snapshots from about:memory.

Now, I can tell you that my brother's system is almost identical to mine, except that I know I do not have IcedTea installed for him (though his Flash is at the same level as mine: Version: 0.4.1.0 - Shockwave Flash 11.5 r502).

I will uninstall IcedTea, now that I've rolled the JVM back to GA3, and see what I get (this may impact results in #65, as well).

[Lewis Rosenthal]

Time to close this. After finally spending some quality time with my desktop, the issue was apparently lack of attention (read: INIs got way too big, and this is the same old out-of-memory issue we see with a lot of things). So, a good cleaning, a diet, a few hundred KB off of their enormous size, and I no longer see this issue (thankfully).

Sorry for the bother!