Boost C++ Libraries: Ticket #8034: Warning fixes in multi_index

attachment set

Tue, 12 Feb 2013 09:28:05 GMT

attachment → multi_index.diff

Joaquín M López Muñoz — Tue, 12 Feb 2013 09:56:35 GMT

Hi Franz,

I fail to see how the patch (which just changes some variables names n, comp to np, comp_) can silent any warning --unless you're using preprocessor symbols with the same name or something. Can you describe these warnings in more detail?

Tue, 12 Feb 2013 10:02:09 GMT

Hi,

it's a warning about a class member that gets hidden by a parameter or variable name (with clang on Mac / XCode 4.5).

Such issues might lead to hard-to-find bugs in the code, therefore this warning is enabled in our projects.

Joaquín M López Muñoz — Tue, 12 Feb 2013 11:25:58 GMT

OK, I see. I'll fix this. Can you please run the whole Boost.MultiIndex test suite so as to locate all similar problems with the codebase? (There must be more of these, for instance in <boost/multi_index/hashed_index.hpp>)

Thank you,

attachment set

Tue, 12 Feb 2013 12:11:54 GMT

attachment → boost_hashed_index.txt

Tue, 12 Feb 2013 12:20:26 GMT

I've attached a compile log with the warnings. I didn't supply a fix, because these are exactly the kind of issues for which the warning is made for. For the non-maintainer (and maybe even for you) it's really hard to see which places need to be fixed and which not :

e.g. const CompatibleHash& hash,const CompatiblePred& eq)const

{

std::size_t buc=buckets.position(hash(k));

I guess the class CompatibleHash has an operator () which creates a hash and I need to fix it here. But if this is not the case and hash(aValue) is a member of the class hashed_index, it probably shouldn't be fixed.

I didn't dig into the details, but maybe you understand that such code is *really* hard to read (and even harder to fix) for non-maintainers.

Joaquín M López Muñoz — Tue, 12 Feb 2013 15:27:46 GMT

The warnings seems to correspond to test_composite_key.cpp only. Can you run the whole test suite so that we're sure we collect all warnings?

Wed, 13 Feb 2013 10:51:12 GMT

The compilation of the other test files produced only duplicates of these warnings / warnings in other libraries (like the test framework). I'm pretty sure that we've collected all warnings in your code.

status changed; resolution set

Joaquín M López Muñoz — Fri, 15 Feb 2013 06:51:51 GMT

status new → closed
resolution → fixed

(In [82897]) fixed #8034

Joaquín M López Muñoz — Fri, 15 Feb 2013 06:55:38 GMT

Franz, can you locally check that this clears your warnings? Thank you.

Fri, 22 Feb 2013 07:50:51 GMT

Hi,

I've compiled our libraries against the current boost trunk, no multi_index warnings any longer.

Thank you very much!