On Sun, Jan 23, 2000 at 05:09:37PM -0500, Roland McGrath wrote: > > I've been testing my Kaffe build under the Hurd. Most simple java programs > > run fine, however there still are plenty ones that dont. While trying to > > debug test cases, one program once run consistently make gnumach issue a > > "panic: zalloc" and crash. I've been able to place the panic call in > > gnumach-1.2/kern/zalloc.c line 486 (source taken from the debian archive). > > However if run under gdb, this only gives a SEGV in a seemingly unrelated > > place. Any clue to what's causing the problem? > > If you can produce a small case that causes this panic, please send it in. > This is an indication of the kernel running out of memory in a bad way that > it should not usually be doing. > > > BTW, if gnumach hangs, is this a bug? > > Most likely.
I don't know about a small test case, since I don't really have a good idea what's causing it. But I've finally scrambled some time to put together a binary distribution of kaffe for the Hurd. You can get it at ftp://scol.concordia.ca/pub/kaffe-bin.tar.bz2, it should run out of the box if you unpack it directly in the /usr/local directory, or you could upack it in a separate directory but would have to change a few scripts. Also get the files regression-classes.tar.bz2 and all.failed from the same directory. The former contains precompiled test programs and the latter the list of the test programs that failed. The ones that produce a "panic: zalloc" error are UncaughtException, NullPointerTest and HashtableTest1. So, any other Java enthusiasts, try to d/l and install kaffe to see if you can find any more bugs in it or in the Hurd. Igor
