valgrind-users

[Valgrind-users] Print summary before exec

[François-Xavier C. <fx....@ya...>]

Hi,

I have a program that calls exec without forking. I would like to run
that program under valgrind and get a summary report for the code before
the call to exec. I don't want to trace the exec'ed program, so the
--trace-children option does not do what I want.

Is there a way to have valgrind output the error/leak summary before
exec (without tracing after exec)?

Thanks,
François-Xavier

Re: [Valgrind-users] Print summary before exec

[Paul F. <pj...@wa...>]

> De : "François-Xavier Carton via Valgrind-users"

> Hi, I have a program that calls exec without forking. I would like to run that program under valgrind
> and get a summary report for the code before the call to exec.
> I don't want to trace the exec'ed program, so the --trace-children option does not do what I want.
> Is there a way to have valgrind output the error/leak summary before exec (without tracing after exec)? 

Hi Francois-Xavier

Sure, this is possible. There are two possible things that you can do

1. Use vgdb. To do this you run Valgrind and your test app in one terminal and in a second
terminal you run gdb which connects to the first. In the gdb terminal you can issue
most of the usual commands like 'break' and 'continue'. Before the exec you can do
'monitor leak_check full'. See
https://www.valgrind.org/docs/manual/mc-manual.html#mc-manual.monitor-commands
for details.

2. Use client requests. To do this you need to make some small changes to your code.
You need to include memcheck.h, then just before your exec() insert

VALGRIND_DO_LEAK_CHECK;

For more details, see also the manual.

A+
Paul

Re: [Valgrind-users] Print summary before exec

[François-Xavier C. <fx....@ya...>]

On Thu, Jun 25, 2020 at 09:49:35PM +0200, Paul FLOYD wrote:
> 
> > De : "François-Xavier Carton via Valgrind-users"
> 
> > Hi, I have a program that calls exec without forking. I would like to run that program under valgrind
> > and get a summary report for the code before the call to exec.
> > I don't want to trace the exec'ed program, so the --trace-children option does not do what I want.
> > Is there a way to have valgrind output the error/leak summary before exec (without tracing after exec)? 
> 
> Hi Francois-Xavier
> 
> Sure, this is possible. There are two possible things that you can do
> 
> 1. Use vgdb. To do this you run Valgrind and your test app in one terminal and in a second
> terminal you run gdb which connects to the first. In the gdb terminal you can issue
> most of the usual commands like 'break' and 'continue'. Before the exec you can do
> 'monitor leak_check full'. See
> https://www.valgrind.org/docs/manual/mc-manual.html#mc-manual.monitor-commands
> for details.
> 
> 2. Use client requests. To do this you need to make some small changes to your code.
> You need to include memcheck.h, then just before your exec() insert
> 
> VALGRIND_DO_LEAK_CHECK;
> 
> For more details, see also the manual.
> 
> A+
> Paul
> 

Hi Paul,

Thanks! I will try to go with 2., because I would like to automate that.

François-Xavier