Apache OpenOffice (AOO) Bugzilla – Issue 84209
Letter wizard broken
Last modified: 2013-08-07 14:42:49 UTC
just compiled src238 and letter wizard got broken. looks like a regression and it's maybe the same like http://www.openoffice.org/issues/show_bug.cgi?id=83670 it's x86_64 ArchLinux. the wizard doesn't come up with neither gcj or Sun java. output from command line: [andyrtr@workstation64 extra]$ soffice-dev com.sun.star.lang.IllegalArgumentException: at com.sun.star.bridges.jni_uno.JNI_proxy.dispatch_call(Native Method) at com.sun.star.bridges.jni_uno.JNI_proxy.invoke(JNI_proxy.java:183) at $Proxy19.invoke(Unknown Source) at com.sun.star.wizards.common.Resource.getStringList(Resource.java:76) at com.sun.star.wizards.letter.LocaleCodes.<init>(LocaleCodes.java:15) at com.sun.star.wizards.letter.LetterWizardDialogImpl.initializeNorms (LetterWizardDialogImpl.java:895) at com.sun.star.wizards.letter.LetterWizardDialogImpl.startWizard (LetterWizardDialogImpl.java:125) at com.sun.star.wizards.letter.CallWizard$WizardImplementation.trigger (CallWizard.java:95)
Reassigned to HI.
I can confirm the bug here on my i686 src680_m240 build also using gcj. This seems to be a regression since it works in OOo 2.3.1
Please check if it's related to issue 84680.
Doesn't seem related to issue 84680 but it seems the exact same as issue 83670 although that one is marked fixed so maybe the bug is still there?
Exactly, this is a duplicate to issue 83670, the is currently planned to be fixed for OOo2.4. *** This issue has been marked as a duplicate of 83670 ***
.
Reopening this one. Issue 83670 was already processed in a CWS and shall NOT be used for further code changes.
Reassigned to OS. Set dependency to "OOo 2.4 blocker" issue 84957.
Adding CCs from closed issue 83670. Set keyword "regression".
sba, this isn't only a 64bit issue. it happens on my 32bit Linux installation as well.
SBA: It does NOT occur in these 32bit installations: OOH_m5 (SUN build 9264) on SUSE (KDE) with SUN JRE 1.6.0. OOH_m6 (SUN build 9268 on Fedora (Gnome) with "Free Software Foundation" JRE 1.4.2. SBA->ht990332: The 32/64 bit thingie is not the key to the core of the problem. Please give further data of your installation, thank you.
sba: i built from source using package-format=native gcc 4.2.2 problem occurs with both builds using gcj and sun's java.
I bet I know what this is, probably simple. i.e. cmc->all who can reproduce it: Are you using ooo-build or other wise using a patch which modifies svtools/inc/svtools/svtools.hrc e.g. most likely candidate of the popular patches is the gstream integration one which adds to resources into that file above the resource whose id is hardcoded into letter.java. If you are, then just try this little patch to bump the no by two, and if it works then some ooo-build user should add it to the e.g. gstreamer set of patch
Created attachment 51372 [details] patch for those that have patched svtools.inc only
cmc: Yeah, I can confirm that this patch works. Adding to ooo-build. Now the interesting question is why the Sun autotest did find it, too ;-)
@rene: I don't know if a test for the letter wizard exists, but I assume it wasn't found mainly because the problem doesn't exist on builds made by Sun
mba: no, read again (the original issue 83670). The issue *WAS* found by your autotests :)
Really? What I can see there is that the autotest found an issue with the letter wizard. This one was fixed already on the CWS (it was a bug in the java code). After that people reported they "still have this bug" but no one using Sun Builds was able to reproduce it. That sounds like there is a second bug in non-Sun builds that just creates the same results. And thanks to Caolan this bug seems to be fixed now also!
closing. I think we can agree that hardcoding a resource id into the java code is fragile, but I assume its non-trivial to come up with a better solution.
SBA-> rene: As you recently told me: "The ability to read..." I will try again: (1) The AutoTest found A PROBLEM OF A NOT WORKING LETTER WIZARD in the CWS "Langstatusbar". Therefore I wrote [CWS langstatusbar] in the title. (1a) My description includes "...can not be closed normally thereafter. Neither File-Exit nor the window closer work..." Now if this tiny detail is different, my QA nose tells me that this is a different issue than the one I found in the CWS. (2) My description starts with "Found by Autotest "f_updt_wizard_letter.bas" This means that that AutoTest threw an error while running on that CWS. (3) My description proceeds with " Is OK in master SRC680_m232" This means that that AutoTest threw NO error while running on Master build m232. (4) Later in that issue, I wrote "Verified in re-built and re-synced (to m237) CWS langstatusbar." This means that that the very same AutoTest threw NO error while running on re-synced (to m237) CWS langstatusbar. (5) The AutoTest shows no findings in the Master builds after langstatusbar. Summary: Sometimes a partly similar issue can have a different root cause. A developers analysis of a root cause helps much more than shouting and escalating.
SBA-> rene: As you recently told me: "The ability to read..." I will try again: > (1) The AutoTest found A PROBLEM OF A NOT WORKING LETTER WIZARD in the CWS > "Langstatusbar". Therefore I wrote [CWS langstatusbar] in the title. Yes, as was the case here too (although not in cws langstatusbar) > (1a) My description includes "...can not be closed normally thereafter. > Neither File-Exit nor the window closer work..." Now if this tiny detail is > different, my QA nose tells me that this is a different issue than the one I > found in the CWS. It isn't. > (2) My description starts with "Found by Autotest "f_updt_wizard_letter.bas" > This means that that AutoTest threw an error while running on that CWS. > (3) My description proceeds with " Is OK in master SRC680_m232" > This means that that AutoTest threw NO error while running on Master build m232. That per se is not an argument, as it might have worked by chance. But I overread this, sorry. > Summary: Sometimes a partly similar issue can have a different root cause. ACK. > A developers analysis of a root cause helps much more than shouting and > escalating. Correct. But you are not a developer. And I did not shout anywhere in the conversation. I was just defending that this is no 64-bit only problem and what I referenced was your initial comment. That the root cause was something else is something else. You were right this time, but I have often seen issues affecting upstream bogusgly which then either got forgotten or needed to be reopened. I feared this would have happened here, too. And since there will be no 2.4.1.... (and I had bad mood at tha time anyway...)
i have to reopen this issue: on i686 ArchLinux it helped to use letter-wizard- resource-id.diff patch. but on x86_64 ArchLinux the wizards keeps to fail opening. tested with m7. any idea?
andyrtr: works for me on x86_64 Debian. (in fact, x86_64 is my main platform, so anything I try will be tried on x86_64 first)
removing gstreamer support made it working again. so it can be closed.
close again
and "close": re-open on pain of death.