Menu
▾
▴
#76 segv possibly relating to named blocks
We have some trouble which seems to relate to named blocks.
I enclose a test case which, for me, fails with a
segmentation fault running 0.4.5 (I filed against cvs
because there is not yet a 0.4.5 group on sourceforge)
I was unable to remove the define or the include
without losing the segmentation fault.
I've included the output of valgrind, which shows an
invalid read and hopefully demonstrates a problem even
if the segv isn't reproducible.
Please let me know if you need any more from me!
Thanks
Ed
ed.spittles at artofsilicon.com
source files and logs including valgrind log
Logged In: YES
user_id=512771
I have created a 0.4.5 group now (thanks for letting me
know about that) in which this bug report is included
now. I will take a look at this problem later today.
Trevor
Logged In: YES
user_id=512771
Ed,
I have attached a tarball containing the files that fix this
bug. Just copy it to the covered-0.4.5/src directory and
unzip/untar. Since you seem to be finding several issues at
the moment, I will attempt to fix them and let you give them
a try before I create a new stable release version. If this
method is undesirable/unuseful to you, let me know how you
would like to receive fixes going forward.
Thanks,
Trevor
Tarball containing some source files that fix this issue (and a few other memory leak issues)
Logged In: NO
I'm happy to test with updates like this, or I could access
CVS if that is better for you.
I've applied both your tarball updates.
I can confirm that this testcase no longer segv's - and
indeed I can now run on a larger design which was previously
giving trouble.
Many thanks for your rapid response (and indeed for the
years of work on the tool!)
Cheers
Ed
Logged In: YES
user_id=512771
I am changing this bug report to the pending as the
described problem seems to be fixed but is not yet available
as a downloadable release. The fix for this problem will be
available in the 0.4.6 stable release.