Red Hat Bugzilla – Bug 435048
pfmon using uninitialized value on ia64
Last modified: 2008-04-08 08:43:24 EDT
Description of problem:
The latest pfmon package fails to build on ia64:
dear_hist_ia64.c: In function 'dear_hist_process_samples_normal':
dear_hist_ia64.c:189: error: 'csmpl' is used uninitialized in this function
<.. snip ..>
void *hash_desc = csmpl->data, *data;
csmpl really never is set to anything before we set a pointer to one of it's
fields. Looking at where hash_desc is used I see that we really do need this to
be initialized to something.
Since csmpl is later initialized as:
csmpl = &sdesc->csmpl;
I _assume_ what was meant for the hash_desc initialization was to initialize
hash_desc after csmpl is set. Please confirm this is what is intended (it
compiles but I don't know enough about pfmon to really test it).
Also, there was one other much more trivial case of using an uninitialized
variable in this patch:
inst_hist_ia64_old_smpl.c: In function 'inst_hist_show_results':
inst_hist_ia64_old_smpl.c:311: error: 'ns' may be used uninitialized in this
I will post a patch that fixes both issues.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
Created attachment 296007 [details]
patch to fix ia64 build
Created attachment 301349 [details]
Avoid using bogus/null values
I don't have an ia64 machine running f9 to check this out on. Could you verify
that the attached patch resolves this problem? If it does, I will go ahead and
put it in the package.
The pfmon-3.3-0.080225.3.fc9 source rpm includes patch. Could you verify that it
builds without error on ia64 fedora?
Your patch addresses the csmpl bit but not the part about "ns" being
uninitialized. See the build log here:
A patch has been checked in to address the initialization of ns. Assuming no
other lurking ia64 build issues, pfmon-3.3-0.080225.4.fc9 should build.
The latest version does indeed build cleanly on ia64. Thanks!