Bug #1369: SIGSEGV in libcrystax.so - lrintf - CrystaX NDK - Issue tracker

Actions

Copy link

Bug #1369

closed

SIGSEGV in libcrystax.so - lrintf

Added by Jakob R. almost 8 years ago. Updated almost 8 years ago.

Status:

Closed

Priority:

High

Assignee:

Dmitry M.

Category:

libcrystax

Target version:

11.0.0

Start date:

04/21/2016

Due date:

% Done:

100%

Estimated time:

CPU Architecture:

arm

Host OS:

OS X

Toolchain:

Android version:

4.1 (android-16)

CrystaX NDK Version:

10.3.1

Description

Hi,
my Android gradle project builds fine and starts to run on my Android device, however I always get a SIGSEGV at a certain line.
This project is a multi-platform project and all the other platforms (OS X, iOS) work fine. It also worked with Google's NDK r10d.
I'm also using OpenCV 3.1 and compiled it successfully with your cmake toolchain with the newest daily build.
I'm building for armeabi-v7a.
Is this a bug in Crystax or am I doing something wrong here?

I get the following stack trace:

04-21 17:33:23.909 4775-4858/de.formigas.lottoscanner A/libc﹕ Fatal signal 11 (SIGSEGV), code 2, fault addr 0x9d536ffc in tid 4858 (AsyncTask #2)
04-21 17:33:24.011 191-191/? A/DEBUG﹕ *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** ***
04-21 17:33:24.011 191-191/? A/DEBUG﹕ Build fingerprint: 'google/razor/flo:6.0.1/MMB29K/2419427:user/release-keys'
04-21 17:33:24.011 191-191/? A/DEBUG﹕ Revision: '0'
04-21 17:33:24.011 191-191/? A/DEBUG﹕ ABI: 'arm'
04-21 17:33:24.011 191-191/? A/DEBUG﹕ pid: 4775, tid: 4858, name: AsyncTask #2 >>> de.formigas.lottoscanner <<<
04-21 17:33:24.011 191-191/? A/DEBUG﹕ signal 11 (SIGSEGV), code 2 (SEGV_ACCERR), fault addr 0x9d536ffc
04-21 17:33:24.036 191-191/? A/DEBUG﹕ r0 9d53700c r1 00000001 r2 88000000 r3 88000000
04-21 17:33:24.036 191-191/? A/DEBUG﹕ r4 9d53700c r5 43aa0000 r6 9d637f88 r7 02010000
04-21 17:33:24.036 191-191/? A/DEBUG﹕ r8 00000020 r9 b4d4abf4 sl cccccccd fp 00000000
04-21 17:33:24.036 191-191/? A/DEBUG﹕ ip b37b7cb8 sp 9d537008 lr b37a5b8f pc b3768c9c cpsr 60070030
04-21 17:33:24.038 191-191/? A/DEBUG﹕ backtrace:
04-21 17:33:24.039 191-191/? A/DEBUG﹕ #00 pc 0001dc9c /data/app/de.formigas.lottoscanner-1/lib/arm/libcrystax.so
04-21 17:33:24.039 191-191/? A/DEBUG﹕ #01 pc 0005ab8b /data/app/de.formigas.lottoscanner-1/lib/arm/libcrystax.so (lrintf+10)

Related issues 1 (1 open — 0 closed)

Actions

Copy link

Updated by Dmitry M. almost 8 years ago

Category set to libcrystax
Assignee set to Dmitry M.
Priority changed from Normal to High

Jakob Raible wrote:

Is this a bug in Crystax or am I doing something wrong here?

Well, if it crashes, it looks as a bug. Could you provide some minimal example to reproduce the problem? I could say more detailed then what's wrong.

Actions

Copy link

Updated by Jakob R. almost 8 years ago

Dmitry Moskalchuk wrote:

Jakob Raible wrote:

Is this a bug in Crystax or am I doing something wrong here?

Well, if it crashes, it looks as a bug. Could you provide some minimal example to reproduce the problem? I could say more detailed then what's wrong.

Thanks for the quick reply!
I digged a little deeper and it definitely has to do with some implicit float to int casting. lrintf seems to be causing the error somehow, as the stack trace already suggests.
An example which leads to a crash is the following:

cv::Point2f pf = cv::Point2f(1.0, 2.0);
cv::Point p = pf;

I'm still busy creating a minimal example, but this might already help you!
Thanks for your efforts!

Actions

Copy link

Updated by Dmitry M. almost 8 years ago

Jakob Raible wrote:

I'm still busy creating a minimal example, but this might already help you!

OK, I confirm that lrintf cause the crash on minimal example:

#include <math.h>
#include <stdio.h>

int main()
{
    long v = lrintf(23.45f);
    printf("v=%ld\n", v);
    return 0;
}

Call stack points inside fegetenv, which is implicitly called from lrintf. It's unclear yet why this happens, but I'm going to figure out what's wrong and fix it.

Thank you for reporting it!

Actions

Copy link