jasonmolenda wrote:

> Just to record that I thought about it, I agree with not trying to make the 
> names architecture neutral.
> 
> For example one might say that because range watchpoints on mips and aarch64 
> have overlapping functionality (no pun intended) that we could report that we 
> have "range" as a type for either of them. This leads to a situation where 
> the functionality that doesn't overlap becomes much harder to handle and 
> detect.
> 
> So I agree with `aarch64-mask` instead of simply `mask`, as an example.
> 
> If we later want to (and kinda already do) group these by general 
> functionality, we can do that in lldb in the algorithms you added for working 
> out the watchpoint resources.
> 
> (and sod's law, someone will propose to add 
> `aarch64-bas-but-this-one-thing-is-different` to the architecture one of 
> these days :) )

Yeah I thought about specifying a more structured name, like "ARCH-STYLE" so 
we'd have "x86_64-mask" in addition to "aarch64-mask", but my guess is a lot of 
targets will work fine with "power of 2, up to sizeof(void*)" and never need to 
adjust this.  Ted Woodward pointed out that the PowerPC watchpoints are quite 
different, so if we support that target we'll definitely need an additional 
algorithm for sure.

https://github.com/llvm/llvm-project/pull/80376
_______________________________________________
lldb-commits mailing list
lldb-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits

Reply via email to