hi, On Mon, Oct 20, 2008 at 03:30:14PM +0300, Eugene V. Lyubimkin wrote: > P.S. No one answered on my investigation message before :(
BTW, I am not part of the APT team. I tried to address this bug after I read http://lists.debian.org/debian-devel-announce/2008/10/msg00000.html > Though, I think, severity of this bug can be downgraded (imho, to important) > if it show itself rarely. This is something that either the APT team or the d-release team should decide though. > Secondly, you have used wrong APT option. 'APT::Cache-Limit', not > 'APT::Cache-File' is actual option. Sorry. > I totally disagree with tag 'unreproducible'. I can reproduce it 100% on my > machine with small amount of apt cache. > Try to do any valuable (with updates) > update with '-o APT::Cache-Limit=20000' and see "dynamic MMap ran out of > room...". I finally managed to trigger a bug. By changing /etc/apt/sources a lot of times, and by using apt-get -o APT::Cache-Limit=100 update I got a segfault. I then recompiled APT 0.7.14 using DEB_BUILD_OPTIONS=noopt,nostrip fakeroot debian/rules build and I run it as root as follows cd apt-0.7.14/build export LD_LIBRARY_PATH='./bin:/lib/' gdb --args apt-get -o 'APT::Cache-Limit=100' update so I could get this backtrace Program received signal SIGSEGV, Segmentation fault. 0x000000370987d00b in memcpy () from /lib/libc.so.6 (gdb) bt #0 0x000000370987d00b in memcpy () from /lib/libc.so.6 #1 0x00007fe5018a5d76 in pkgCacheGenerator (this=0x7fff09b220d0, pMap=0x23aee20, Prog=0x7fff09b22a90) at pkgcachegen.cc:60 #2 0x00007fe5018a71b5 in pkgMakeStatusCache ([EMAIL PROTECTED], [EMAIL PROTECTED], OutMap=0x7fff09b22b60, AllowMem=false) at pkgcachegen.cc:872 #3 0x00007fe50189b40e in pkgCacheFile::BuildCaches (this=0x7fff09b22b60, [EMAIL PROTECTED], WithLock=true) at cachefile.cc:68 #4 0x000000000040d745 in ?? () #5 0x00007fe50185db15 in CommandLine::DispatchArg (this=0x7fff09b23340, Map=0x7fff09b23230, NoMatch=true) at contrib/cmndline.cc:337 #6 0x000000000040a5f5 in ?? () #7 0x000000370981e1a6 in __libc_start_main () from /lib/libc.so.6 #8 0x0000000000406289 in ?? () #9 0x00007fff09b23458 in ?? () #10 0x000000000000001c in ?? () #11 0x0000000000000004 in ?? () #12 0x00007fff09b23bac in ?? () #13 0x00007fff09b23bbd in ?? () #14 0x00007fff09b23bc0 in ?? () #15 0x00007fff09b23bd5 in ?? () #16 0x0000000000000000 in ?? () Note that, to trigger it, you have to force a rebuild of APT caches; to this end, I changed the mirrors in /etc/apt/sources list. (If I run 'apt-get update' succesfully once, then apt-get -o 'APT::Cache-Limit=100' update will not crash, unless I change /etc/apt/sources list again). I tried to give a look into the source code, but unfortunately I don't know it enough to understand what is going on. All this said, the bug I am triggering is somewhat artificial. If I run 'apt-get update' simply, it all works fine in my case. So please provide some more info. Do you have an example situation where /etc/apt/sources.list does not work with 'apt-get update' (with no options - and also please check what you have in /etc/apt/apt.conf ) ? a. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
