Svadev mailing list

[Santosh Nagarakatte <santosh.nagarakatte AT gmail.com>]

Thanks for the information and patch.

Santosh

On Mon, May 21, 2012 at 6:37 PM, Will Dietz 
<willdtz AT gmail.com>
 wrote:
> On Mon, May 21, 2012 at 4:41 PM, Will Dietz 
> <willdtz AT gmail.com>
>  wrote:
>> On Mon, May 21, 2012 at 8:51 AM, John Criswell 
>> <criswell AT illinois.edu>
>>  wrote:
>>> On 5/18/12 5:07 PM, Santosh Nagarakatte wrote:
>>>> Hi all,
>>>>
>>>> I wanted to know if there is an automatic mechanism to generate the
>>>> bitcode after every intermediate pass that safecode runs  when I use
>>>> the -fmemsafety flag.
>>>
>>> In response to your first question, no, I don't know of a way to do
>>> this.  You might ask on cfe-dev and see if they know of one.
>>>
>>
>> "-mllvm -print-after-all" is what you're looking for, although this
>> will cause it to print out the bitcode between *every* pass being
>> executed, including those added by safecode.
>>
>> (There's also "-mllvm -print-after" and "-mllvm -print-before" that
>> take lists of passes, but I don't know how they work and would
>> requiring enumerating all the passes anyway).
>>
>> Unfortunately on my build this causes safecode's clang to crash (but
>> this doesn't happen on my builds of clang-3.0 and clang-3.1).  Not
>> sure what's happening there, but perhaps it will work for you.
>>
>
> Because I was curious, I tracked down the cause of the crash I was seeing.
>
> Cherry-picking upstream llvm's r149480 fixes the issue.  Not sure why
> my build of vanilla clang-3.0 didn't suffer from this same behavior,
> but oh well.
>
> Patch attached that applies to LLVM 3.0.  Apply with "patch -p1 < 
> patchfile".
>
> ~Will



-- 
Santosh G Nagarakatte,
PhD Student,
Computer and Information Science Department
University of Pennsylvania,
Philadelphia-19104
http://www.cis.upenn.edu/~santoshn