Bug#595921: Future unblock: frama-c/20100401+boron+dfsg-5
Package: release.debian.org
Severity: normal
User: release.debian.org@packages.debian.org
Usertags: freeze-exception
Hi release team,
I received a tiny patch from upstream which fixes a performance bug
that could lead to a "stack overflow" error (a crash) during large
analyses. The patch is as follows:
Index: src/value/kf_state.ml
===================================================================
--- src/value/kf_state.ml (revision 9760)
+++ src/value/kf_state.ml (revision 9761)
@@ -44,7 +44,8 @@
try Value.is_accessible (Kstmt (Kernel_function.find_first_stmt kf))
with Kernel_function.No_Statement -> false)
-let mark_as_called kf = Is_Called.add kf true
+let mark_as_called kf = Is_Called.replace kf true
(* ************************************************************************* *)
(** {2 Callers} *)
There is no bug report for this issue (yet) since I got the patch
directly from upstream.
Would it be ok for upload an updated Frama-C package with this change
only? Uploading a new Frama-C would require rebuilding Why as well on
all architectures because it provides a plugin for Frama-C which
contains a hash of some internal modules of Frama-C (that's needed by
OCaml).
And yes, the runtime dependencies of Why are somehow broken since it
doesn't reflect the need of at least the version of Frama-c which was
used during the build. I intended to work on that but didn't find
time. It will be fixed for Wheezy.
Regards,
--
Mehdi Dogguy
-- System Information:
Debian Release: squeeze/sid
APT prefers testing
APT policy: (990, 'testing'), (500, 'testing-proposed-updates'), (500, 'unstable'), (1, 'experimental')
Architecture: amd64 (x86_64)
Kernel: Linux 2.6.32-5-amd64 (SMP w/4 CPU cores)
Locale: LANG=en_US.utf8, LC_CTYPE=en_US.utf8 (charmap=UTF-8)
Shell: /bin/sh linked to /bin/dash
Reply to: