Renato Golin via llvm-dev
2016-Mar-16 10:09 UTC
[llvm-dev] [cfe-dev] Bugs with arm-none-eabi
On 15 March 2016 at 21:07, Abe Clements via cfe-dev <cfe-dev at lists.llvm.org> wrote:> Is anyone working on these bugs? If not where would I start to fix them?Hi Abe, I'm copying llvm-dev, since this is an LLVM matter, not a Clang one. If the bug is not assigned to anyone, you can assume no one is working on it. If you assign it to yourself and no one complains, you can be sure of it. :)> https://llvm.org/bugs/show_bug.cgi?id=25495This is not so trivial, and the answer might just be that we won't support it by design. But it may not. So what you have to do is to gather all information about macros and their usage in inline assembly, including links to documentations, email threads, etc. Then you need to have a try at implementing it, most likely you'll have to modify ARMAsmParser.cpp, and present your changes with the initial analysis. Why the bureaucracy? Because GNU extensions are riddled with undocumented bugs that got abused, then became features, and now they're "supported", but still undocumented. Being in a different community, we don't have enough visibility on whether they think this is a good hack that must stay, or a bad one that no one has removed yet. Thus, it's quite possible that GNU's behaviour changes from the specific undocumented implementation A to another, equally undocumented B, and people will now fill bugs again on LLVM.> https://llvm.org/bugs/show_bug.cgi?id=25508I've seen movement in the list about a similar case. Check commit r258504 and the fallout, which was covered by the review http://reviews.llvm.org/D17183. May be relevant to this case. Also, I believe ARM was working to clean up all macros, so I'm copying some ARM folks to give you a hand on the state of the macros. cheers, --renato