On Oct 21, 2007, at 11:24 AM, Jeff Johnson wrote:

>
> On Oct 21, 2007, at 9:37 AM, Bernhard Rosenkraenzer wrote:
>
>> Installing any package with current CVS rpm (on Linux, glibc  
>> 2.6.1, gcc 4.3)
>> results in a segfault.
>>
>> The segfault occurs in headerGetExtension(), he->t at the time is
>> RPM_STRING_TYPE, he->p.str is non-NULL, but invalid (probably  
>> uninitialized).
>> tagName(he->tag) at the time of the crash is Sha1header.
>>
>> headerGetExtension is called from addTE() when the crash occurs.  
>> (addTE is
>> called from rpmteNew, which in turn is called from  
>> rpmtsAddInstallElement).
>
> Yup.
>
> Try
> 	cvs up -D "7 days ago"
> this bug is going be a pain to find.
>

Bracketed between 4-5 days ago with these changes:

+    - jbj: start regularizing hae/hme calls.
+    - jbj: drill rpmTagCount everywhere.
+    - jbj: change hPTR_t to hRET_t in ABI, revert to const void * in  
API.
+    - jbj: drill hPTR_t through prototypes. hRET_t * refactor next.
+    - jbj: drill rpmTagCount through rpmdb/* files.
+    - jbj: preliminaries to encapsulate headerSprintf goop.
+    - rpm.org: Fix broken doxygen comment.
+    - rpm.org: Don't unnecessarily include stdio.h
+    - jbj: extract verifyscript when available. more to do ...

Digging, only 2311 lines in the diff, so I should have a fix soonishly.

73 de Jeff