Hi Pedro,
It came from this line:
https://github.com/apache/incubator-mxnet/blob/master/src/initialize.cc#L44
It brings us no information other than a "segmentation fault : 11". This came
from internal to send an issue if there is a build failure.
Thanks,
Qing
On 6/6/18, 11:17 AM, "Pedro Lar
No, I did not capture details.
On Wed, Jun 6, 2018 at 11:16 AM, Pedro Larroy
wrote:
> Naveen, do you have more information about why the signal handler was
> causing issues for the scala package? This was also failing on my PR
> validation which included backtraces for crashes in CI.
>
> On Wed,
Naveen, do you have more information about why the signal handler was
causing issues for the scala package? This was also failing on my PR
validation which included backtraces for crashes in CI.
On Wed, Jun 6, 2018 at 6:38 PM, Anton Chernov wrote:
> The problem still persists with make build, so
The problem still persists with make build, so no correlation with the
build system.
2018-06-06 15:49 GMT+02:00 Naveen Swamy :
> I am using make to build
>
> On Wed, Jun 6, 2018 at 6:47 AM, Anton Chernov wrote:
>
> > Yes, we have checked that as well, but did not help in our case. I've
> > check
I am using make to build
On Wed, Jun 6, 2018 at 6:47 AM, Anton Chernov wrote:
> Yes, we have checked that as well, but did not help in our case. I've
> checked out a commit close to 1.1 where it was still working and built it
> with the new ci scripts (using cmake build for armv7) and it failed
Yes, we have checked that as well, but did not help in our case. I've
checked out a commit close to 1.1 where it was still working and built it
with the new ci scripts (using cmake build for armv7) and it failed very
similar. It seems that the problem might be in the way we are building the
library
By the way it turned out the problem was when we used USE_SIGNAL_HANDLER along
with a combination of flags, try removing signal handler and see if it works
> On Jun 6, 2018, at 12:28 AM, Anton Chernov wrote:
>
> Unfortunately, I think this is the same behaviour that we're observing on
> Raspber
Thanks Anton. This is proving to be a very painful and difficult crash to
debug and diagnose, since the stack is not usable from GDB. I don't know if
there's any other option rather than bisect on these cases.
Pedro.
On Wed, Jun 6, 2018 at 9:28 AM, Anton Chernov wrote:
> Unfortunately, I think
Unfortunately, I think this is the same behaviour that we're observing on
Raspberry Pi's. Currently we are bisecting the release to find the breaking
commit to have an idea what exactly is broken.
What I can say for now that this failure is not deterministic (on RPi's)
and the library import to py
Could you compile with debug symbols or get a core file? From this output
is not clear why the crash is happening.
On Sun, May 27, 2018 at 10:04 AM, Naveen Swamy wrote:
> Hi,
> I am working to publish MXNet-Scala package to maven and encountering an
> issue when trying to build with openblas/mkl
Hi,
I am working to publish MXNet-Scala package to maven and encountering an
issue when trying to build with openblas/mkl/apple. This is on both the
master and the 1.2.0 branch? Can some one help with this.
make scalapkg fails when it calls the MXNet backend to get all the APIs ?
can someone help h
11 matches
Mail list logo