DavidSpickett wrote:

> My understanding is that all of the tag interpretation and error handling 
> happens on the LLDB client side, so these aspects are already tested by the 
> Linux tests. I essentially only re-implemented the existing qMemTags packet 
> in debugserver and added integration tests for the most important use cases.

Ok but for my own education can you tell me whether the following situation can 
actually happen on your systems?
```
tagged page
untagged page
tagged page
```
With this layout on Linux I may end up reading across boundaries so parts of my 
memory read will have tags and parts won't. I wonder if you have a more "per 
process" rather than "per page" model.

Also I bring this up because I'm not running those Linux tests with any 
regularity. If/when we have MTE hardware to hand, it'll happen but I'm not 
pushing for it myself.

If the above scenario cannot happen in your systems, or you have your own 
integration testing anyway, this is not a problem for you. But of course make 
your own judgements here.

If you did want to test that stuff, you might be able to port the Linux tests 
but I wouldn't try too hard to do so. There's going to be enough differences I 
bet that mean the tests would be clearer staying separate.

https://github.com/llvm/llvm-project/pull/160952
_______________________________________________
lldb-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits

Reply via email to