Comments on: Compiler Warnings http://www.airs.com/blog/archives/159 Ian Lance Taylor Wed, 19 Nov 2008 06:07:57 +0000 http://wordpress.org/?v=1.5.2 Comment on Compiler Warnings by: Ian Lance Taylor http://www.airs.com/blog/archives/159#comment-11531 Wed, 19 Mar 2008 01:33:07 +0000 http://www.airs.com/blog/archives/159#comment-11531 That is an interesting idea. gcc generally aims -O0 at a combination of easy debugging plus fast compile times. We could have an option to abandon that. There would still be the problem that the warnings would change with new gcc releases, though. That is an interesting idea. gcc generally aims -O0 at a combination of easy debugging plus fast compile times. We could have an option to abandon that.

There would still be the problem that the warnings would change with new gcc releases, though.

]]> Comment on Compiler Warnings by: ncm http://www.airs.com/blog/archives/159#comment-11488 Tue, 18 Mar 2008 06:54:01 +0000 http://www.airs.com/blog/archives/159#comment-11488 It's good to know that's acknowledged as a bug. I don't care whether warnings come from the front end or the optimizer. What would be helpful is if we could compile -O0 for comprehensible debugging, and run the optimizer at -O∞ for the warnings but otherwise discard the results. I would run it that way all the time, when developing. It’s good to know that’s acknowledged as a bug.

I don’t care whether warnings come from the front end or the optimizer. What would be helpful is if we could compile -O0 for comprehensible debugging, and run the optimizer at -O∞ for the warnings but otherwise discard the results. I would run it that way all the time, when developing.

]]> Comment on Compiler Warnings by: Ian Lance Taylor http://www.airs.com/blog/archives/159#comment-11440 Mon, 17 Mar 2008 18:31:47 +0000 http://www.airs.com/blog/archives/159#comment-11440 The issue with the order of warning options is a bug. It's already better in gcc 4.2.3 and 4.3.0. See http://gcc.gnu.org/PR32102. I agree that your example of spurious warnings is a problem. Unfortunately that is the kind of problem which is hard to fix if warnings is only emitted by the frontend. The issue with the order of warning options is a bug. It’s already better in gcc 4.2.3 and 4.3.0. See http://gcc.gnu.org/PR32102.

I agree that your example of spurious warnings is a problem. Unfortunately that is the kind of problem which is hard to fix if warnings is only emitted by the frontend.

]]> Comment on Compiler Warnings by: ncm http://www.airs.com/blog/archives/159#comment-11386 Sun, 16 Mar 2008 07:57:37 +0000 http://www.airs.com/blog/archives/159#comment-11386 My problem with Gcc warnings is that Order Matters. E.g., "-Wstrict-overflow=5 -W" is the same as "-W". IMHO "-W" should never reduce the level of any warning level. The choices above are quibbles in the face of such counterintuitive behavior. Uninitialized-variable warnings are a big problem, as noted by some Linux kernel developers. Spurious warnings, for cases like "if (A) b=c; d(); if (A) e(b);" are especially annoying because if you silence them by initializing b with a dummy value, you wipe out a future, smarter compiler's ability to detect real errors. My problem with Gcc warnings is that Order Matters. E.g., “-Wstrict-overflow=5 -W” is the same as “-W”. IMHO “-W” should never reduce the level of any warning level. The choices above are quibbles in the face of such counterintuitive behavior.

Uninitialized-variable warnings are a big problem, as noted by some Linux kernel developers. Spurious warnings, for cases like “if (A) b=c; d(); if (A) e(b);” are especially annoying because if you silence them by initializing b with a dummy value, you wipe out a future, smarter compiler’s ability to detect real errors.

]]>