Small note on static analyzer, includes and build times...

[Albert Thuswaldner]

Hi Michael,

On Thu, Mar 1, 2012 at 12:28, Albert Thuswaldner
<albert.thuswaldner at gmail.com> wrote:
> Hi Stephan and Michael,
>
> On Wed, Feb 29, 2012 at 19:12, Michael Meeks <michael.meeks at suse.com> wrote:
>>
>>        Well - the pre-compiled headers would make -tons- of includes look
>> useless anyway, though (I thought) we had decided to get rid of them as
>> an under-maintained and bad job anyway ;-)
>>
>>        Have you tried acting on this lot:
>>
>> Looking for useless includes in header files...
>>        AccessibleCell.hxx --> global.hxx (useless)
>>        AccessibleCellBase.hxx --> global.hxx (useless)
>>        AccessiblePreviewHeaderCell.hxx --> global.hxx (useless)
>>        AccessibleTableBase.hxx --> global.hxx (useless)
>>        ...
>>
>>        And turning it into some patches to remove those headers & seeing if it
>> compiles nicely ? if so it'd be nice to include that no doubt.
>>
>
> Sure, I can do that, I'll give it a spin and see if it pops...

To get back to you on this, I have now done a test run where I have
gone through a couple (about 20) of random selected files from the
ones flagged by the analyze script.

The conclusion is that there are too many false positives generated
with this method atm. I did not find any real useless includes in the
sample.

Analyzing the includes marked as useless in more detail all (from the
tested sample) provide custom data type definitions which _are_  used
in the header file. So this relation seems to be overlooked some how
by the script.

I've checked the doxygen output (i.e the generated html), in those
files this relation is actually captured. This leaves me to believe
that the error is either in perlmod or the script itself. I do get
warning messages when running the script, but I have not analyzed them
in detail. Perl does not come natural to me, that's why I shamelessly
CC'ing Thomas Collerton, who seems to be the perl-guy around here. ;)

So Thomas, would you be interested to help out with some perl debugging? :)

If you do want to help, I can send you the necessary files (they break
the 500k limit of this ML)

on perlmod:
http://www.stack.nl/~dimitri/doxygen/perlmod.html

----
On a general note, it would be great to have one tool up and running
so we, as a start, at least could quantify how large the problem with
useless includes really is. I'm willing to make an effort on this and
have already checked another tool cppclean
http://code.google.com/p/cppclean/, without success.

Next, would be to check iwyu
http://code.google.com/p/include-what-you-use/. To get started with
this it would be great to know the status
regarding LLVM/clang, can it be used build master (without too much
hassle)?

I've also checked what Mozilla has been up to in this area:
https://bugzilla.mozilla.org/show_bug.cgi?id=557565
https://bugzilla.mozilla.org/show_bug.cgi?id=634839

https://bugzilla.mozilla.org/show_bug.cgi?id=430328

Rather long email, hope you didn't get bored.

/Albert