Hi Wei-hsing,
This will be taken up in the IBIS Quality Task Group meeting tomorrow. If you
have test case files available they should be sent to Bob Ross.
Mike
From: "Michael Huang" <Michael.Huang@xxxxxxxxxx>
To: ibis-macro@xxxxxxxxxxxxx
Sent: Monday, September 17, 2018 2:52:22 PM
Subject: [ibis-macro] Linux ibischk6.1.4 crashes on duplicated symbol?
Hi,
I would like to report a case which the latest ibischk6.1.4 crashes (on Linux).
The simplified test case is like this:
-) There is an IBIS file with two models, each of the model contains its own
AMI models (.so files)
-) These AMI models may be from same vendor. Though they are for Tx and Rx
respectively and names of the .so are different, part of the source codes they
were built from may be the same or share some common libraries (say "boost"
libraries, but versions may be different).
IbisChk crashes (segmentation fault) when checking this ibs file. The
interesting thing is that if I comment out the [Algorithmic Model] for one of
them...one at a time, then ibischk works fine and the .so file passes the check
for each of them! IbisChk crashes only when both two different .so files are
present in the .ibs file.
I am thinking there may be symbol collisions or global variable scoping issue.
Maybe the checker's library loading mechanism has set some limitations (e.g.
via RTLD_DEEPBIND or RTLD_LOCAL flags etc). In either case, the checker should
not crash... at least it should exist gracefully. Is this a bug or are there
some extra requirements not stated in the spec. which model vendors need to
follow in order for the .so files to be demangled correctly (when they are
present in the same .ibs file)?
Thanks for your comments.
Regards,
Wei-hsing
-----------------------------------------------
SPISim: [ [ http://www.spisim.com/ ;| www.spisim.com ] ], Free tool: SPILite
EDA expertise in Signal, Power Integrity and SIMulation
-----------------------------------------------