Hello everyone,
I'm developing an application for several months using Orocos 2.3.1 with the
CORBA transport layer. Actually, for Each call to a remote operations I
get a memory leak of about 200 bytes in the TLSF real-time memory pool (see
bug 889 [1]). For example, with the allocated default size of 20k, I can't
make more than 100 calls to remote operations before the application crashes
(std::bad_alloc exception from TLSF). I'm curious if anyone is actually
using the CORBA protocol and whether they also detect this memory leak? I'd
like to hear about your experience with this, and your configuration (ie,
version of RTT) in order to detect if the problem is unique to my setup. So,
in brief, could someone reproduce the problem?
Thank you!
Philippe Hamelin
A call to the Orocos 2.x users of the CORBA transport layer
On Friday 23 September 2011 14:09:26 Philippe Hamelin wrote:
> Hello everyone,
>
> I'm developing an application for several months using Orocos 2.3.1 with
> the CORBA transport layer. Actually, for Each call to a remote operations
> I get a memory leak of about 200 bytes in the TLSF real-time memory pool
> (see bug 889 [1]). For example, with the allocated default size of 20k, I
> can't make more than 100 calls to remote operations before the application
> crashes (std::bad_alloc exception from TLSF). I'm curious if anyone is
> actually using the CORBA protocol and whether they also detect this memory
> leak? I'd like to hear about your experience with this, and your
> configuration (ie, version of RTT) in order to detect if the problem is
> unique to my setup. So, in brief, could someone reproduce the problem?
>
> Thank you!
>
> Philippe Hamelin
>
> [1] http://bugs.orocos.org/show_bug.cgi?id=889
You've seen the patch by now I suppose. I'm sorry this couldn't be fixed
earlier for you since it was <1h of work, especially with the excellent test
cases. The recent projects I'm working on are less 'Orocos support' and more
'application development', which leaves less time to fix bugs that don't affect
me directly. As always, keep pinging the mailing list... it does work !
Just make sure that your SendHandle/SendHandleC are scoped properly, and all
cleanup will be done.
Peter
A call to the Orocos 2.x users of the CORBA transport layer
On Sep 23, 2011, at 08:09 , Philippe Hamelin wrote:
> Hello everyone,
>
> I'm developing an application for several months using Orocos 2.3.1 with the CORBA transport layer. Actually, for Each call to a remote operations I get a memory leak of about 200 bytes in the TLSF real-time memory pool (see bug 889 [1]). For example, with the allocated default size of 20k, I can't make more than 100 calls to remote operations before the application crashes (std::bad_alloc exception from TLSF). I'm curious if anyone is actually using the CORBA protocol and whether they also detect this memory leak? I'd like to hear about your experience with this, and your configuration (ie, version of RTT) in order to detect if the problem is unique to my setup. So, in brief, could someone reproduce the problem?
While I can't speak for v2, I know that in v1 we do _not_ get this problem. I don't believe there have been any changes in TLSF itself, so that points to the CORBA side as well.
Personally, I'd be happy to look at this, if I could ever get RTT v2 to build and pass all its tests on my development platform ... sigh ... :-(
Cheers
S
A call to the Orocos 2.x users of the CORBA transport layer
2011/9/23 S Roderick <kiwi [dot] net [..] ...>
> On Sep 23, 2011, at 08:09 , Philippe Hamelin wrote:
>
> > Hello everyone,
> >
> > I'm developing an application for several months using Orocos 2.3.1 with
> the CORBA transport layer. Actually, for Each call to a remote operations I
> get a memory leak of about 200 bytes in the TLSF real-time memory pool (see
> bug 889 [1]). For example, with the allocated default size of 20k, I can't
> make more than 100 calls to remote operations before the application crashes
> (std::bad_alloc exception from TLSF). I'm curious if anyone is actually
> using the CORBA protocol and whether they also detect this memory leak? I'd
> like to hear about your experience with this, and your configuration (ie,
> version of RTT) in order to detect if the problem is unique to my setup. So,
> in brief, could someone reproduce the problem?
>
> While I can't speak for v2, I know that in v1 we do _not_ get this problem.
> I don't believe there have been any changes in TLSF itself, so that points
> to the CORBA side as well.
>
>
I also have a legacy system running on 1.x and I don't have this problem.
This only occurs since 2.x.
> Personally, I'd be happy to look at this, if I could ever get RTT v2 to
> build and pass all its tests on my development platform ... sigh ... :-(
>
>
I would be happy to have a second opinion from you. I have attached two
complete test cases to the bug in order to easily test the memory leak by
two different ways.
Philippe
A call to the Orocos 2.x users of the CORBA transport layer
On Sep 23, 2011, at 08:43 , Philippe Hamelin wrote:
> 2011/9/23 S Roderick <kiwi [dot] net [..] ...>
> On Sep 23, 2011, at 08:09 , Philippe Hamelin wrote:
>
> > Hello everyone,
> >
> > I'm developing an application for several months using Orocos 2.3.1 with the CORBA transport layer. Actually, for Each call to a remote operations I get a memory leak of about 200 bytes in the TLSF real-time memory pool (see bug 889 [1]). For example, with the allocated default size of 20k, I can't make more than 100 calls to remote operations before the application crashes (std::bad_alloc exception from TLSF). I'm curious if anyone is actually using the CORBA protocol and whether they also detect this memory leak? I'd like to hear about your experience with this, and your configuration (ie, version of RTT) in order to detect if the problem is unique to my setup. So, in brief, could someone reproduce the problem?
>
> While I can't speak for v2, I know that in v1 we do _not_ get this problem. I don't believe there have been any changes in TLSF itself, so that points to the CORBA side as well.
>
>
> I also have a legacy system running on 1.x and I don't have this problem. This only occurs since 2.x.
>
> Personally, I'd be happy to look at this, if I could ever get RTT v2 to build and pass all its tests on my development platform ... sigh ... :-(
>
>
> I would be happy to have a second opinion from you. I have attached two complete test cases to the bug in order to easily test the memory leak by two different ways.
I have a number of bugs registered that I need Peter's help on first. Until I can trust that RTT v2 actually works for me (by passing its tests), I'm not sure I can be of much help.
S