kcachegrind-callgrind — Discussion around callgrind usage and development

Re: [Kcachegrind-callgrind] Using Kcachegrind with XDebug profile results

[francois m. <fm...@gm...>]

Thanks Josef,

I knew I couldn't be that wrong ;)

Let's wait for an update then !


2011/2/14 Josef Weidendorfer <Jos...@gm...>

> On Friday 11 February 2011, francois mazerolle wrote:
> > Hi,
> >
> > I'm trying to use Kcachegrind with the XDebug results.
> > Basically, it seem to work, but I'm facing the problem where I think that
> > KCachegrind doesn't do all the links -- I think -- he should do.
>
> See https://bugs.kde.org/show_bug.cgi?id=256425
>
> Josef
>



-- 
Francois Mazerolle;
- Membre W3Québec <http://www.w3qc.org/membres/individus/>
- Membre OWASP (Quebec City <http://www.owasp.org/index.php/Quebec_City>)
- Administrateur CyberCity2034 <http://www.cybercity2034.com/> & Projet
Tholus <http://www.tholus.org/> (Gagnant
FTW<http://www.webnotwar.ca/ftw/2010-winners/>
)
- Rédacteur developpez.net <http://www.developpez.net/forums/u58499/fmaz/>
- Projet JQuery Image WaterMark<http://plugins.jquery.com/project/imgwatermark>

Re: [Kcachegrind-callgrind] Using Kcachegrind with XDebug profile results

[Josef W. <Jos...@gm...>]

On Friday 11 February 2011, francois mazerolle wrote:
> Hi,
> 
> I'm trying to use Kcachegrind with the XDebug results.
> Basically, it seem to work, but I'm facing the problem where I think that
> KCachegrind doesn't do all the links -- I think -- he should do.

See https://bugs.kde.org/show_bug.cgi?id=256425

Josef

[Kcachegrind-callgrind] Using Kcachegrind with XDebug profile results

[francois m. <fm...@gm...>]

Hi,

I'm trying to use Kcachegrind with the XDebug results.
Basically, it seem to work, but I'm facing the problem where I think that
KCachegrind doesn't do all the links -- I think -- he should do.

In the Call Graph tab, I often have no more than 2 or 3 items.

I've posted a stackoverflow post here:
http://stackoverflow.com/questions/4971684/kcachegrind-having-difficulties-to-link-php-xdebug-calls

And some kcachegrind screenshot here:
http://img46.imageshack.us/img46/2226/kcachegrind123.png

Basically, my application start from index.php, then require a
frontcontroller.php file.
The front controller do a bunch of things, so I guess I should -- at least
-- have a 3 level call graph is I start by the main:
{Main} -> require frontcontroller.php -> anything (but not nothing).

I've also noticed that some require are in double, with the value (0) in the
Called column.
I'm I doing something wrong here ?

Actually the results are hard to interpret as everything is separated.



-- 
Francois Mazerolle;
- Membre W3Québec <http://www.w3qc.org/membres/individus/>
- Membre OWASP (Quebec City <http://www.owasp.org/index.php/Quebec_City>)
- Administrateur CyberCity2034 <http://www.cybercity2034.com/> & Projet
Tholus <http://www.tholus.org/> (Gagnant
FTW<http://www.webnotwar.ca/ftw/2010-winners/>
)
- Rédacteur developpez.net <http://www.developpez.net/forums/u58499/fmaz/>
- Projet JQuery Image WaterMark<http://plugins.jquery.com/project/imgwatermark>

Re: [Kcachegrind-callgrind] bugs and questions

[Josef W. <Jos...@gm...>]

Dear adu,

On Wednesday 28 January 2009, adu wrote:
> First of all I will ask a question.
> 
> Where and how I can get the latest build of kcachegrind tool if:
> 
> My Linux OS can't be updated up to latest release due to some admin's
> restrictions.
> 
> On site http://kcachegrind.sourceforge.net/html/Home.html
> <http://kcachegrind.sourceforge.net/html/Home.html%20exists%20only%200.4.6>
> /  exists only 0.4.6 version of this tool.
> 
> But I know that in the latest KDE4.2 (package kdesdk) you have improved
> version of this tool.

KCachegrind is currently developed in the KDE repository. The current version
there uses KDE4/Qt4. The version in KDE 4.2 should work quite reasonable with
KDE4 (in contrast to KDE4.0/KDE4.1). However, there were not really new features,
just a bit of polisihing and fixing of bugs introduced with the port to the
KDE4/Qt4 libraries.

There is a need for a command line version, which can write back files after
filtering/merging/..., as well as an interest to use parts of KCachegrind from pure
Qt applications (e.g. to embed parts into valkyrie, the Qt frontend for valgrind
from valgrind.org, or to make a plugin for Qt creator, the new Qt IDE).
Therefore, KCachegrind currently is split into libraries, one containing all the
non-GUI stuff (internal data model, import filter), and one containing the pure
visualization views; however, the API is not fixed yet.

Neverless, you can already compile a pure Qt4 version: checkout from KDE SVN
the trunk/KDE/kdesdk/kcachegrind directory, go into src/qcachegrind and compile
with "qmake; make", run with "./qcachegrind". No KDE libs needed, just Qt4.4
with devel package. See this as example; "qcachegrind" definitly is a very
bad name for the tool ;-)

This will be made available as independent source packages from kcachegrind.sf.net.

> But have I can extract only this tool from that package??

For the KDE4 version, you need to checkout trunk/KDE/kdesdk, remove the
"add_subdirectory()" lines in the CMakeLists.txt apart from kcachegrind,
and run "cmake .; make; make install".

> And some words about the bug. Maybe you have fixed it already.
> ...

No, not yet. This issue is very known, and I have even a patch from
someone here. Aside from relayout of the tooltip, or provided algorithms
to shorten the symbol, the ultimate plan is to provide a hook, ie. a
optional custom script where all symbols are piped trough before display.

Josef

[Kcachegrind-callgrind] bugs and questions

[adu <ad...@na...>]

Hello all!!

 

First of all I will ask a question.

Where and how I can get the latest build of kcachegrind tool if:

My Linux OS can't be updated up to latest release due to some admin's
violations.

On site http://kcachegrind.sourceforge.net/html/Home.html
<http://kcachegrind.sourceforge.net/html/Home.html%20exists%20only%200.4.6>
this exists only 0.4.6 version of this tool.

But I know that in the latest KDE4.2 (package kdesdk) you have impoved
version of this tool.

But have I can extract only this tool from that package??

Regards, Anton
_______________________________________________________________

The content of this message, together with any attachments, are intended
only for the use of the individual or entity to which they are addressed and
contains information that is confidential and could be a trade secret of
Nangate. If you are not the intended recipient, you are hereby notified that
any dissemination, distribution, or copying of this message, or any
attachment, is strictly prohibited. If you have received this message in
error, please notify the original sender by return E-mail and delete this
message, along with any attachments, from your computer. Thank you.

 

[Kcachegrind-callgrind] bugs and questions

[adu <ad...@na...>]

 

Hello all!!

 

First of all I will ask a question.

Where and how I can get the latest build of kcachegrind tool if:

My Linux OS can't be updated up to latest release due to some admin's
restrictions.

On site http://kcachegrind.sourceforge.net/html/Home.html
<http://kcachegrind.sourceforge.net/html/Home.html%20exists%20only%200.4.6>
/  exists only 0.4.6 version of this tool.

But I know that in the latest KDE4.2 (package kdesdk) you have improved
version of this tool.

But have I can extract only this tool from that package??

 

And some words about the bug. Maybe you have fixed it already. 

This tool is supporting ToolTips mechanism as a way to help users to read
dotted lines of text.

For ex. when I want to read full name of function in column 'Function' in
'Flat Profile' view - 

I just have to move a mouse cursor on the desirable function and wait a
little - and ToolTip will bring up.

BUT some functions have STL containers as an arguments - and in this case
ToolTip became so big - 

that I can't read it all (((

You have to create formatted ToolTips - otherwise this useful mechanism is
useless. 

There is no any sense to read something like 

ass::PORT_KINDS const, myProxyClass::PORT_TYPES>,
std::pair<myProxyClass::PORT_KINDS const, myProxyClass::PORT_TYPES>&,
std::pair<myProxyClass::PORT_KINDS const,
myProxyClass::PORT_TYPES>*>(std::_Rb_tree_iterator<std:

 

Thanks in advance)))

 

P.S. to moderators - pls, remove my previous mail from 28/01/9, 12:05. It
was not complete ((

Re: [Kcachegrind-callgrind] can not get partial dumps of profile data

[Josef W. <Jos...@gm...>]

Hi,

On Thursday 09 March 2006 16:07, Giovanni Sartoni wrote:
>  I am using callgrind 0.9.11  (valgrind-2.2.0) to profile a C++ test
> application that 
> comprises 4 steps, I would like to selectively collect profile data only when
> the application runs the method:
>
> IncrementalTester::step4(void)
> ... 
> callgrind
> -v
> --log-file=reflTestIncrement

This redirects Callgrinds log messages; if you wanted to redirect the profile
data to another file, use "--base=...".

> --num-callers=20

This option has no meaning for Callgrind, that's for memcheck's backtraces
in its error output.

> --simulate-cache=no

Not needed, is default.

> --cacheuse=no

Not needed, is default.
 
> --dump-before=IncrementalTester::step4 
> --dump-after=IncrementalTester::step4

I suppose this is wrong. For C++, callgrind is using demangled C++ symbols,
which include the full signature. So you have to say

 "--dump-before=IncrementalTester::step4(void)"
or
 "--dump-before=IncrementalTester::step4*"


> --collect-atstart=no
> --toggle-collect= IncrementalTester::step4

Should work with "...step4*".

> option --dumpat=<funct> is not valid

That is obsoleted by "--dump-before".


Hope this helps.
Josef

[Kcachegrind-callgrind] can not get partial dumps of profile data

[Giovanni S. <gsa...@fu...>]

Hi

 I am using callgrind 0.9.11  (valgrind-2.2.0) to profile a C++ test
application that

comprises 4 steps, I would like to selectively collect profile data only =
when

the application runs the method:

=20

IncrementalTester::step4(void)

=20

The application is compiled with gcc 2.95.3.

I have tried the following:

=20

callgrind

-v

--log-file=3DreflTestIncrement

--num-callers=3D20

--simulate-cache=3Dno

--cacheuse=3Dno

--dump-before=3DIncrementalTester::step4

--dump-after=3DIncrementalTester::step4

incremental_tester.exe parameters.txt

=20

but I still get one single profile dump file callgrind.out.xxxxx, with =
the
whole run in it.

I also tried

=20

callgrind

-v

--log-file=3DreflTestIncrement

--num-callers=3D20

--simulate-cache=3Dno

--cacheuse=3Dno

--collect-atstart=3Dno

--toggle-collect=3D IncrementalTester::step4

incremental_tester.exe parameters.txt

=20

but still no success. Furthermore the option --dumpat=3D<funct> is not =
valid
for this version.

Any help to solve this is highly appreciated.

=20

Best regards

=20

Giovanni Sartoni

=20

[Kcachegrind-callgrind] Fwd: Callgrind callgrind-0.10.1 broken configure

[Clint O. <ol...@ic...>]

----- Forwarded message from Clint Olsen <ol...@ic...> -----

Hi Josef:

I'm having trouble getting configure to pass on callgrind.  Currently it is
checking for $prefix/bin/valgrind for the executable, but I'm setting
exec_prefix in the install, and this is tripping up configure's ability to
find my valgrind install.

I'm trying to patch this now, but configure is such a freaking mess in
general that it's a royal pain.  If I get a patch I'll send it to you.  You
may have a preferred method of repairing the problem.

-Clint

----- End forwarded message -----

Re: [Kcachegrind-callgrind] Re: kcachegrind reporting .so files to be without debug info

[Josef W. <Jos...@gm...>]

On Saturday 17 September 2005 18:13, Steinar Bang wrote:
> >>>>> Josef Weidendorfer <Jos...@gm...>:
> >
> > This should be a problem in Valgrinds debug reader then. If you run
> > memcheck, are there any errors so that you can verify that this
> > problem is in fact Valgrinds?
>
> Do I need to run the entire procedure (which took the entire night)?
> Or is it enough that I just run the program a little with memcheck?

Better: Start
 valgrind --trace-symtab=yes yourProg &> log
for a few seconds and check log to see if the missing symbols had already 
problems being read in from the debug info section.s loading the symbols.

Josef

[Kcachegrind-callgrind] Re: kcachegrind reporting .so files to be without debug info

[Steinar B. <sb...@do...>]

>>>>> Josef Weidendorfer <Jos...@gm...>:

> This should be a problem in Valgrinds debug reader then. If you run
> memcheck, are there any errors so that you can verify that this
> problem is in fact Valgrinds?

Do I need to run the entire procedure (which took the entire night)?
Or is it enough that I just run the program a little with memcheck?

[Kcachegrind-callgrind] Re: OK to subscribe kcachegrind-callgrind to gmane?

[Josef W. <Jos...@gm...>]

On Thursday 15 September 2005 08:15, Steinar Bang wrote:
> Hi!
>
> Would it be OK if I asked that the kcachegrind-callgrind mailing list be
> subscribed to the mail<->news gateway at http://gmane.org/?

I am fine with it.

Thanks,
Josef

[Kcachegrind-callgrind] Re: kcachegrind reporting .so files to be without debug info

[Josef W. <Jos...@gm...>]

On Thursday 15 September 2005 08:42, you wrote:
> [Emailing directly to you since the sourceforge mailing list looks
>  like it doesn't get much use]

That probably would be the perfect discussion to start using the mailing 
list...

> That is: it finds the source for the main() function, and it finds the
> source for all inline code, but it can't find the source for any of
> the code in the .so files.  It says eg.:
>
>  There is no source available for the following function:
>    '0x3C6BBBC4'
>  This is because no debug information is present.
>  Recompile source and redo the profile run.
>  The function is located in this ELF object:
>    'libsomename.so.1.0'
>
> However I know I compiled that .so (as well as everything else) with
> debug info.  And I'm able to run the program in gdb and set
> breakpoints in functions in that library, and look at variables when I
> stop there.

This should be a problem in Valgrinds debug reader then. If you run memcheck, 
are there any errors so that you can verify that this problem is in fact 
Valgrinds?

> Is there some option I should have run callgrind with?

No.

> Is there some 
> g++ demangling difference between 3.3 and 4.0 (the kcachegrind I'm
> using is linked against libstdc++.so.5 so I guess that means it's
> compiled with gcc 3.3 as well.  Don't know if that means anything?).

KCachegrind does not do any demangling. This is done by VG, and the profile 
data contain the demangled names.

Besides, this only would be the problem if you get mangled symbols. But you 
get no debug info as all, as far as I see.

Josef

>
> Thanx!
>
>
> - Steinar