Mozilla crashes 100% reproduceably when trying to open a (possibly incorrect) mail attachment. To clarify, the browser does not crash but cleanly exits, shutting all windows cleanly, and spits out the following message on the command line. INTERNAL ERROR on Browser End: [0] Initialize. No docbase? System error?:: Resource temporarily unavailable Program exited with code 0377. Since mozilla has exited cleanly, there is NO coredump or crash for us to investigate. The lack of resource message made me this we could be running out of stack, but increasing the shell stack size does NOT help. This is 100% reproduceable on Solaris AND Linux on all nightly builds and the stock mozilla 1.0.

This file should be forwarded as an ATTACHMENT in mozilla. When you recieve it, doubleclick on the README file to show the attachment.

I think that since the attachment i created was from a "Save As" in the message pane window, it probably didn't preserve the headers correctly and may not reproduce the problem clearly. If someone wants me to forward them the file as an attachment, please send me an email dave@0bits.com and i will forward them the file as an attachment. It is only 13K big.

Confirmed: Linux/x86 20020723 (so it's been around a while).

From the error message, it _seems_ to be a Java issue, do you crash when deactivating Java ?

Nope, still happens with Java truned off completely. My Java/Javascript is turned off by default for Mail/News.

confirming crash using build 20020805 on Linux (trunk) + JRE 1.4.1 beta. When double-clicking on the README file (forwarded message), here's what the console show: [...] WARNING: CreateSecureEnv OK, file ProxyJNI.cpp, line 1679 For application/x-java-applet found plugin /usr/lib/java/plugin/i386/ns610/libjavaplugin_oji.so INTERNAL ERROR on Browser End: [0] Initialize. No docbase? System error?:: Resource temporarily unavailable Program exited normally. Current language: auto; currently c [...] This is the standard message when Mozilla loads a page with a Java applet but, here, why is it launching a Java process ?? I don't get any other stack trace (although I'm running a debug build). btw, it doesn't seem to crash build 20020805 on Win2k (trunk).

More comments to comment #6, if I disable Java within Mozilla, I still crash. If I remove the Java plug-in (softlink) from my plugins/ directory, I don't crash. On Win2k, if I open the file README, it fails, and then Mozilla is unusable (zombie), even after quitting Mozilla, it remains in memory (not using QuickLaunch). Weird. I tried to produce testcases but failed for now (empty file called README, didn't crash), I'm trying to reduce the README in size...

Adding plugin folks to the CC, as there's a weird plugin interaction going on here.

why is the plain text file interacting with java? why would jave even be invoked?

hmm, no crash for me. Dave, could you forward the test case attachment to me, please? thanks.

I forwarded the messsage onto Serge. However this is now trivial to reproduce.. 1. Save the 1st attachment above as a text file called "README" - no .txt suffix 2. Forward it to yourself using mozilla mail as an attachment. 3. Select/Read the mail you just recieved with the attachment. 4. Click on the attachment 5 [details]. Bang ! Your Mozilla's dead.

Ok, i did some more testing and i notice that if you add the attachment to your mail as a file called 'README.txt' then it does NOT crash your mail window upon reading the message *and* it gets correctly displayed in the forwarded message. So somehow the file type detection is breaking in the attachments and it doesn't detect the README file as a txt file. It looks like the default file type is being set as JAVA and is trying to be interpreted as such.

Dave thanks, for test case, I can reproduce it now, it looks like plugins issue, reassigning myself for investigation.

similar weird bug: bug 157102 (where Java was totally unrelated).

attachments without extensions are assigned a mime-type of application/x-java-vm dupe of "mail attachment with mime application/x-java-vm crashes Linux" *** This bug has been marked as a duplicate of 141826 ***

changing qa contact to yulian

verified dup