On 12/13/2013 08:41 PM, Josh Stone wrote:
Hi,
I have published a batch of fixes for issues reported by Coverity:
https://sourceware.org/git/gitweb.cgi?p=dyninstfans.git;a=shortlog;h=refs/heads/jistone/coverity
I'd like to get these in before I disappear for the holiday, though I
may have already missed you reviewers departing. :)
Nope, I'm around all week and then out for the next two.
The bulk of these fixes are for UNINIT_CTOR, i.e. POD-typed members that
are not initialized in the constructor or its callees. There are a few
other logic errors fixed too, and a few things that just needed syntax
tweaks to avoiding getting flagged as a possible issue.
There's nothing too scary, I think. I'd like to push this early next
week if there's no review blocker.
Notes:
* Symtab's Object copy constructor: pretty sure it should copy all the
fields if it's going to exist. The secondary point, of course, is that
all Symtabs/Objects/AObjects should be reference counted and (in
principle) should never be copied, so the more accurate fix is to
replace them with private versions etc.
* The linux_kludges assorted fixes look much simpler and cleaner, which
is great, but I also have to ask if these are still needed/used anywhere
(or if it's safe to replace them with POSIX equivalents in some cases).
As a general rule, things in common that are not actually
domain-specific are things that we'd like to get rid of but (at the time
we wrote them) there wasn't a sufficiently cross-platform alternative.
(See also: pdvector. The cruft goes a long way back here.)
* Not 100% positive that all of the fallthroughs that you left as
fallthrough are non-bugs, but in the worst case they're not making
behavior any worse. I'll review those in more detail this afternoon and
see if I find anything that needs fixing.
Thanks,
Josh
_______________________________________________
Dyninst-api mailing list
Dyninst-api@xxxxxxxxxxx
https://lists.cs.wisc.edu/mailman/listinfo/dyninst-api
--
--bw
Bill Williams
Paradyn Project
bill@xxxxxxxxxxx
|