Boost C++ Libraries: Ticket #7949: Level 4 Warnings with VS2005

attachment set

Tue, 29 Jan 2013 12:46:46 GMT

attachment → lexical_cast.hpp.diff

Silence warning C4244 under VS2005

status, milestone changed

Antony Polukhin — Tue, 29 Jan 2013 17:10:49 GMT

status new → assigned
milestone To Be Determined → Boost 1.54.0

status changed; resolution set

Antony Polukhin — Fri, 01 Feb 2013 07:16:35 GMT

status assigned → closed
resolution → wontfix

This is a bug of VS2005, it is fixed in VS2012 (may be also fixed in earlier versions of VS). Working around buggy warnings on non default warning level - I don't think that it is worth doing in lexical_cast.

Fri, 01 Feb 2013 09:58:56 GMT

Replying to apolukhin:

Working around buggy warnings on non default warning level - I don't think that it is worth doing in lexical_cast.

Well, maybe it is not, yes. Admittedly this particular warning is rather braindead under some circumstances (notably more or less unneccessary integer promotion done by the compiler behind the scenes. Short int seems to be treated as a second class citizen here). Anyway, consider:

The problem with warnings coming from headers of course is that they often will popup in every including cpp-file or at every place an certain template instance is used, cluttering the build output and hindering/demotivating programmers fixing their selfmade warnings.
There are companies that implement a zero-warning-policy - they have a reason to do so.
"Tightening the screws" on programmers by increasing the warning level is a useful thing, it often leads the way to uncover hidden, subtle bugs.
Boost itself states they intend to make warning free headers for every supported compiler - with additional warnings enabled (/W4, -Wall, etc.).
Many programmers unfortunately are bound to a certain compiler/compiler version - and to being molested by the warnings it emmits, no matter whether they're buggy or not.

So that's why I submitted this, and the patch is just a #pragma to shut down the warnings in lexical_cast, no line of production code is being changed. But never mind, that's just my personal opinion and everyone's summoned to have their personal opinion.

Best regards

Peter

status changed; resolution deleted

Antony Polukhin — Fri, 01 Feb 2013 11:48:17 GMT

status closed → reopened
resolution wontfix

OK, you convinced me. But I'll disable this warning only for VS2005 (_MSC_VER == 1400). If you could reproduce it on VS2008, I'll also fix it there.

Fri, 01 Feb 2013 13:12:21 GMT

Replying to apolukhin:

OK, you convinced me. But I'll disable this warning only for VS2005 (_MSC_VER == 1400).

Yep, that's fine.

If you could reproduce it on VS2008, I'll also fix it there.

Sorry, but this is a no-op for me! I do not have access to VS2008 (AKA for the time being I'm bound to VS2005 & gcc 4.3.3 (WindRiver version, somehow patched for vxWorks) and may not simply install an express edition of VS2008 or whatever) :-/

Someone else has to check whether this poses a problem in VS2008, too. Of course I could install VS at home for testing purposes - but everybody wants to call it a day eventually, so do I. <irony> You don't want my wife to punish me, do you? </irony> ;->

Best regards

Peter

Antony Polukhin — Sun, 10 Feb 2013 13:59:47 GMT

(In [82807]) Silence some buggy VS warnings (refs #7949)

status changed; resolution set

Antony Polukhin — Sun, 24 Feb 2013 08:31:19 GMT

status reopened → closed
resolution → fixed

(In [83114]) Merge from trunk:

Silence some buggy VS warnings (fixes #7949)
Fix -Wconversion warnings (fixes #2558)
Added support for int128 conversions (fixes #7909)