Sorry, I feel like I have to make a stand here. This whole
story (including earlier patches) is getting incredibly
ridiculous. There are nice warning options in todays
compilers, but if a warning option sucks and makes us
write worse code, then we shouldn't use it. I've wasted
so much time discussing such options, that I could have
spend reviewing and fixing actual bugs instead. IMHO,
some warning options just make things worse.

If we feel like we need to write /* zero initialized */
because we do that implicitly and rely on that, while the
language actually offers a nice and clean way to do that
explicitly, then we failed, desperately. If we want a nice
Clang experience, please let's just disable that ridiculous
warning.

Patch set 7:Code-Review -1

1 comment:

File util/ich_descriptors_tool/ich_descriptors_tool.c:
- Patch Set #2, Line 155: struct ich_descriptors desc = { 0 };
  Ron suggested using static which works well.
  Well, here is how much I remember of the story:
  - empty braces {} are not allowed in C
  - { 0 } is valid C for any aggregate even if it starts with a sub-aggregate (the 0 will just fill the first field of the first sub recursively, all other fields will be filled like static objects)
  - GCC folks made {} and { 0 } exceptions to -Wmissing-field-initializers, otherwise nobody would use that option because people would get mad, writing all the 0s.
  - Clang folks had a patch to make { 0 } an exception too, maybe they rolled it back, idk.

To view, visit change 36946. To unsubscribe, or for help writing mail filters, visit settings.