[cfe-commits] [llvm-commits] Reduce exception handling use in TableGen

[Sean Silva]

>> There is an error in your patch, but I'm not going to tell you where. Good luck ;-)
>
> Maybe I am still too sleepy from the conference, but I don't see any :(

Try applying your patch to master. I think r166382 introduced a conflict.

-- Sean Silva

On Tue, Oct 23, 2012 at 12:07 PM, Joerg Sonnenberger
<joerg at britannica.bec.de> wrote:
> On Mon, Oct 22, 2012 at 10:26:48AM -0700, Jakob Stoklund Olesen wrote:
>>
>> On Oct 22, 2012, at 1:08 AM, Joerg Sonnenberger <joerg at britannica.bec.de> wrote:
>>
>> > Hi all,
>> > the attached patch removes the majority of the exception handling
>> > dependency in TableGen. There is a small regression in the quality of
>> > error messages in one code path of the SetTheory implementation, where
>> > the location of the original expression is lost. This is a bit difficult
>> > to fix since the functions involved can be accessed directly as well and
>> > I am not sure how much value the location adds here.
>>
>> There is an error in your patch, but I'm not going to tell you where. Good luck ;-)
>
> Maybe I am still too sleepy from the conference, but I don't see any :(
>
>> The location-less errors are really annoying, please try to find a way
>> of preserving the location info. It also looks like many of the throws
>> you converted (outside SetTheory) could easily pass a relevant location
>> to PrintFatalError.
>
> Sure, can work on providing that, but it changes the interface.
>
>> Is it possible to get rid of the TGError exception class as well?
>
> I think so. Without throwing exceptions around, it is just a fancy way
> to store message and location.
>
> Joerg
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits