From Bugzilla Helper: User-Agent: Mozilla/4.8 [en] (Windows NT 5.0; U) Description of problem: rpm -ivh --prefix /opt/intel/compiler80 intel-icc8-8.0-38.i386.rpm intel-isubh8-8.0-38.i386.rpm Segmentation fault Operation okay without --prefix option Version-Release number of selected component (if applicable): rpm-4.2.1-0.30 How reproducible: Always Steps to Reproduce: 1. Need a relocatable product 2. install with rpm -iivh --prefix /abc 3. Actual Results: segmenation fault Expected Results: correct execution Additional info:
I am not sure I am see the same problem or not but I get a segment fault when I try installing "old" rpms build on rhl 8 and rhl 9. rpm is 4.2.1-0.30
Something definitely strange here. I tried another ("simple") rpm built on rhl 9 and tried installing it ... no segfault this time, rpm just "hung". After waiting a bit (about a minute), I killed the rpm process, cleanup /var/lib/rpm/, did --rebuilddb (just to be safe), and then tried the install again ... this time with the same package build on FC1 T3 -- it installed fine (ten seconds total). rpm -Uvh for src rpms seems to work fine.
BTW, suggest subject be changed to just "rpm fails with segmentation fault"
Does this problem track with --prefix or not? Seperate bug reports please. It's really hard to sort out 2 problems from the same report.
Gene, would you mind telling us the names of at least some of the packages that are segfaulting for you? Even naming just one or two of the packages would greatly help others attempt to reproduce the problems -- especially since it sounds like you might be seeing segfaults on some packages and freezes on others, if I'm understanding your comments correctly. Posting a separate bug report, as Jeff suggests, would be a great idea too. If it turns out to be the same bug after all, it can be easily marked as a duplicate later, and if it turns out to be a different bug than what Winfrid reported, you'll save everyone else a lot of effort.
OK, I will open another report for my problem. I was going to at first but then thought this was really the same problem. I will create a new report. The packages I started having the problem with were the VMware-workstation-3.2.0 and 4.0.5 packages (these can be downloaded from vmware). I have no idea what release thee were build on. I then tried a package of my own that had been built on RHL8 and has few dependencies. This install hung. A package rebuilt on FC1T3 installed. I am going to try the failing package again. I can make this package available if you need it. I have also had up2date hang trying to install today's updates from rawhide.
Then again, I guess I will not open a new report ... the problem has "disappeared". I just successfully removed my package and installed the one that had hung earlier. I then successfully installed VMware-workstation-4.0.5 which had gotten a segfault earlier. I am unclear about the up2date hang at this time. Between the hang/segfault problems and these successful installs, I did a --rebuilddb.
If you see segfaults or other problems again, try running: /usr/lib/rpm/rpmdb_verify /var/lib/rpm/Packages (hopefully I remembered that correctly) and include the output (if any) in your bug report. BTW, if up2date seems to be hanging, try giving it a *lot* of time (e.g., leave it there for half an hour or an hour and see if it's still frozen). Particularly on slower machines or when accessing large yum repos (think rawhide) over slow connections, up2date can be slow and can seem frozen when it's really not. Also, if you were using up2date 4.1.5, try 4.1.7 -- it seems a faster to me.
This error occurred for the same level of rpm on RHEL 3.0 beta, it is gone now, because rpm has now a higher level ( rpm-4.2.1-4.2 ) so you should use at least the same level on fedora. Winfrid
The segmentation fault of rpm with --prefix option fails still with fedora core 1 release. The problem could easily be solved with a higher version of rpm . Could you please provide rpm-4.2.1-4.x for fedora in the fedora updates. Thanks, Winfrid
OK, there's at least 4 problems here, all different. Please make a new bug report with a reproducible test case. The version of rpm used and a ptr (i.e. URL) to the package are absolutely essential for me to attempt to reproduce. Winfrid: Where is intel-icc8-8.0-38.i386.rpm?
Because I wrote, that you can produce this error with any rpm which is relocable I did not attach the rpm. The rpm is quite big so I looked for a smaller one. Instructions to install ( to produce the error ) : rpm -ivh --prefix /opt/intel/compiler80 intel-iidb7-7.3.1-87.i386.rpm Winfrid PS: I tried to upload the rpm, but I am not allowed to transfer a file > 1 mb, please advise me, or download Intel Compiler 8.0 ( either fortran or c ++ ) from www.intel.com/software/products/global/eval.htm
Created attachment 96967 [details] rpm file to reproduce the error Finally I found a smaller rpm-file to reproduce the error : [root@wv2-fedora l_cc_p_8.0.055]# !r rpm -ivh --prefix /opt/intel/compiler80 intel-isubh8-8.0-44.i386.rpm Segmentation fault While rpm command without --prefix works : [root@wv2-fedora l_cc_p_8.0.055]# rpm -ivh intel-isubh8-8.0-44.i386.rpm Preparing... ########################################### [100%] 1:intel-isubh8 ########################################### [100%] Regards, Winfrid
Created attachment 97009 [details] rpm file to reproduce the error Finally I found a smaller rpm-file to reproduce the error : [root@wv2-fedora l_cc_p_8.0.055]# !r rpm -ivh --prefix /opt/intel/compiler80 intel-isubh8-8.0-44.i386.rpm Segmentation fault While rpm command without --prefix works : [root@wv2-fedora l_cc_p_8.0.055]# rpm -ivh intel-isubh8-8.0-44.i386.rpm Preparing... ########################################### [100%] 1:intel-isubh8 ########################################### [100%] Regards, Winfrid
I am seeing this problem as well. If I pass the "--prefix" flag, RPM crashes with a seg fault. Without the "--prefix" flag, it succeeds. Unfortunately, I can't send you the RPM in question as it is part of our internal codebase. I believe it was generated on a machine running red hat 7.2 with rpm version 4.0.3. Is there a short term fix for this problem? I am kind of dead-in-the-water for development otherwise. [b-5-x dev-x86 04:32 PM] rpm -hiv --test --nodeps --ignorearch --prefix $TOOLROOT --dbpath /home/trees/local/b-5-x/toolroot/var/lib/rpm /home/trees/rpms/b-5-x/latest/dev-x86-tools/toolroot/tivotools-tools_ism-default-1.0-131854.i386.rpm Segmentation fault lauterbrunnen:/home/trees [b-5-x dev-x86 04:32 PM] rpm -hiv --test --nodeps --ignorearch --dbpath /home/trees/local/b-5-x/toolroot/var/lib/rpm /home/trees/rpms/b-5-x/latest/dev-x86-tools/toolroot/tivotools-tools_ism-default-1.0-131854.i386.rpm Preparing... ########################################### [100%] lauterbrunnen:/home/trees [b-5-x dev-x86 04:33 PM]
Reproduced the seg fault with the rpm attached to comment 14. Funny thing is I issued $ rpm -iv --prefix /opt/intel/compiler80 intel-isubh8-8.0-44.i386.rpm without adding "sudo". Still see a segfault. Attaching core dump.
Created attachment 98427 [details] Core file for rpm -iv --prefix
Any ETA on a fix?
Hi, I am currently on holiday, so I cannot check this item - could you please check if this error exists on core 1.9x. As I reported, this is a very old bug, but it has be corrected on rhel 3.0 and rh 9.0 Winfrid Please check also fedora core 1 for x86_64
Dell is also seeing this issue on RHEL3. If you take a RHEL3 gold install and then upgrade to all RHN latest packages (pre-Update2) our upcoming systems management driver RPM gets a segmentation fault when attempting to install it. If the rpm is rebuilt to not use prefix, the rpm does not segfault.
I'm not able to get this to replicate on a current RHEL 3 system with the attached RPM: [root@dhcp170 tmp]# rpm -ivh --prefix /opt/intel/compiler80 intel-isubh8-8.0-44.i386.rpm Preparing... ########################################### [100%] 1:intel-isubh8 ########################################### [100%] [root@dhcp170 tmp]# rpm -ql intel-isubh8 /opt/intel/compiler80 /opt/intel/compiler80/bin /opt/intel/compiler80/bin/uninstall.sh /opt/intel/compiler80/substitute_headers /opt/intel/compiler80/substitute_headers/libio.tar.gz [root@dhcp170 tmp]# rpm -q rpm rpm-4.2.1-4.4 Has anyone seen this particular package break for them in testing on RHEL 3?
Hi, I cheked in the meantime fedora 1.91 for x86_64 - there is no problem, so I suppose that there will be no problem in the upcoming fedora core 2. In case of rhel 3.0 for ipf I saw this problem only in preproduction version. Winfrid Tschiedel
here's a fix for this problem: http://www.mail-archive.com/openpkg-dev%40openpkg.org/msg09531.html
FWIW, the openpkg fix is insuffficient. Otherewise this is fixed in rpm-4.4.2 and later afaik. Reopen if 4.4.2 does not solve the preoblem.