Not that I really expect a lot of help on this, but . . . .
We've got a fairly large (~500,000 lines) Java code base which we want to build
and deliver on multiple platforms ("build once, run everywhere" you say? Read
on). We've run into a problem on DEC.
For the record, we're using Compaq's 1.3.0-1 version of the jdk, ant version
1.2, and the OS level is 5.1.
The problem seems to occur whenever a "Runtime.getRuntime().exec" gets invoked.
This means whenever we invoke the <exec> task and also in a homemade task I
created. For reasons I don't understand, it also occurs when the <copy> task is
invoked. Also, it only happens when I've handed ant a fairly large target (for
example, if I hit the build.xml file at the top of my source tree, I get
multiple errors. On the other hand, if I hit each build.xml at the bottom of by
build tree one at a time, no errors occur).
The error is a sig_signo [11]: SIGSEV 11* segmentation violation. In every
case, a look at the table of processes shows that the main ant process is
sleeping, and a new <java> process has been started and is hung. Killing off
the zombie allows the sleeping ant process to continue with the build. I
believe that when a "Runtime.getRuntime().exec" is executed, a new <java>
process is forked off, but then segfaults and hangs immediately. I have no idea
why this is occuring during a <copy> task.
Any suggestions about why the segfault is occuring during the <copy> task, or at
all?
Thanks,
--dave
