[PATCH] D83152: llvm-nm: add flag to suppress no symbols warning

Fri Feb 26 09:36:13 PST 2021

MaskRay added a comment.

You can use --quiet instead. They accepted my `--quiet` patch: https://sourceware.org/bugzilla/show_bug.cgi?id=27408

In D83152#2559118 <https://reviews.llvm.org/D83152#2559118>, @jhenderson wrote:

> In D83152#2557844 <https://reviews.llvm.org/D83152#2557844>, @MaskRay wrote:
>
>> I don't agree. If the coordination is done first, we have a potential to use `--quiet` to right away. Then we don't need to have a temporary `--no-warning-for-no-symbols` which will eventually be removed (and potential to break users).
>
> Whilst what you are saying is entirely true, the issue is that it unnecessarily (in my opinion) raises the bar on a contribution that is already small and quite simple. Not all developers know or care how to discuss things with the GNU community. Saying that they do makes it more likely that they will simply walk away from contributing what might otherwise be a useful patch. In this case, it's clearly not a code quality issue, nor is it a tricky thing for us to change in the future should some future developer decide they want to discuss the option with GNU/GNU independently go their own way.

Discussing with binutils is case by case. This feature solves a minor insurmountable problem. ("no symbols" can be easily filtered out in the log.)
We would have migration cost that adds a different option first, adds `--quiet` for compatibility, then drops the first option.

>> The `no symbols` has caused us lots of trouble (see the history). I imagined that binutils folks would have some opinions on it. Given the contentious previous discussions, I think it is entirely justified to ask them for an opinion.
>
> Could you point out where these contentious previous discussions were? I looked back through the history, and there didn't seem to be any real issues - it was just about getting the llvm-nm implementation right so that it matches the GNU behaviour, but it's quite possible I missed something, or I might just be misunderstanding you. I don't think difficulties trying to get our implementation to match GNU's are really relevant to this patch, since this is about a new option, so won't impact the default behaviour.

We added `no symbols`, then dropped the diagnostic if there is hidden symbols (e.g. ARM mapping symbols `$a` which are normally not display for arm-*-nm), then changed the output stream from stdout to stderr.

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D83152/new/

https://reviews.llvm.org/D83152