Segmentation fault on WSL (Debian and Ubuntu) about bart

[smyeungx]

Thanks! Forgot to report that NOEXEC_STACK feature actually work very well for me. The same approach is working for nufft, with the following minor modification:

1. in the source code "src/noncart/grid.c"
2. replace the code update(ind, d); with NESTED_CALL(update, (ind, d));

[uecker]

I think it is related to the fact that we now require an executable stack (for nested functions). Apparently WSL does not support this. Oh well, I have to think about whether we can do something about this...

[uecker]

A possible alternative may be to try clang. Put the following in Makefile.local
OMP=0
CC=clang-3.8
CFLAGS+= -fblocks
LDFLAGS+= -lBlocksRuntime
You will loose parallelization with OpenMP though. I had some trouble using other versions of clang.

[smyeungx]

Dear Martin, Thanks for the prompt reply! Let me try and revert. Simon

…

________________________________ From: Martin Uecker <[email protected]> Sent: Monday, August 6, 2018 2:18:29 PM To: mrirecon/bart Cc: smyeungx; Author Subject: Re: [mrirecon/bart] Segmentation fault on WSL (Debian and Ubuntu) (#140) A possible alternative may be to try clang. Put the following in Makefile.local OMP=0 CC=clang-3.8 CFLAGS+= -fblocks LDFLAGS+= -lBlocksRuntime You will loose parallelization with OpenMP though. I had some trouble using other versions of clang. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#140 (comment)>, or mute the thread<https://github.com/notifications/unsubscribe-auth/AoLSoovzLFHwJ4Qsj1cEV0_OqAmb3213ks5uN9-1gaJpZM4VvthK>.

[uecker]

Let me know whether this worked. Then we could document this as a workaround.

[smyeungx]

Hi, After recompilation with the Makefile.local in 0.4.03 with the corresponding flags, zeros command works. simon@DESKTOP-6P96G2J:~/bart-0.4.03$ ./bart zeros 1 1 a The following matlab from website tutorial also works: % Zero padding to make square voxels since resolution in x-y for this % data set is lower in phase-encode than readout smallfov = readcfl('data/smallfov'); smallfov = bart('resize -c 2 252', smallfov); % Direct calibration of the sensitivities from k-space center for SENSE sensemaps = bart('caldir 20', smallfov); % SENSE reconstruction sensereco = bart('pics -r 0.01', smallfov, sensemaps); But doesn’t work for the ecalib % ESPIRiT calibration with 2 maps to mitigate with aliasing in the calibration espiritmaps = bart('ecalib -r 20 -m 2', smallfov); gdb bt the above command gives the following return: bart: /home/simon/bart-0.4.03/src/num/casorati.c:40: void casorati_dims(unsigned int, long *, const long *, const long *): Assertion `dim[i] >= dimk[i]' failed. Thread 1 "bart" received signal SIGABRT, Aborted. __GI_raise (sig=sig@entry=6) at ../sysdeps/unix/sysv/linux/raise.c:51 51 ../sysdeps/unix/sysv/linux/raise.c: No such file or directory. (gdb) bt #0 __GI_raise (sig=sig@entry=6) at ../sysdeps/unix/sysv/linux/raise.c:51 #1 0x00007ffffdc7442a in __GI_abort () at abort.c:89 #2 0x00007ffffdc6be67 in __assert_fail_base (fmt=<optimized out>, assertion=assertion@entry=0x48d5ef "dim[i] >= dimk[i]", file=file@entry=0x48d601 "/home/simon/bart-0.4.03/src/num/casorati.c", line=line@entry=40, function=function@entry=0x48d62c "void casorati_dims(unsigned int, long *, const long *, const long *)") at assert.c:92 #3 0x00007ffffdc6bf12 in __GI___assert_fail (assertion=0x48d5ef "dim[i] >= dimk[i]", file=0x48d601 "/home/simon/bart-0.4.03/src/num/casorati.c", line=40, function=0x48d62c "void casorati_dims(unsigned int, long *, const long *, const long *)") at assert.c:101 #4 0x000000000045584c in casorati_dims (N=<optimized out>, odim=<optimized out>, dimk=<optimized out>, dim=<optimized out>) at /home/simon/bart-0.4.03/src/num/casorati.c:40 #5 0x0000000000448426 in calibration_matrix (calmat_dims=0x7ffffffecc60, kdims=<optimized out>, calreg_dims=0x7ffffffed960, data=0x6aa1f0) at /home/simon/bart-0.4.03/src/calib/calmat.c:34 #6 covariance_function (kdims=<optimized out>, N=<optimized out>, cov=<optimized out>, calreg_dims=0x7ffffffed960, data=0x6aa1f0) at /home/simon/bart-0.4.03/src/calib/calmat.c:161 #7 0x000000000044365e in compute_kernels (conf=0x7ffffffedd70, nskerns_dims=<optimized out>, nskerns_ptr=<optimized out>, SN=288, val=0x7ffffffed4e0, caldims=<optimized out>, caldata=<optimized out>) at /home/simon/bart-0.4.03/src/calib/calib.c:693 #8 0x00000000004452e1 in calone (conf=0x7ffffffedd70, cov_dims=<optimized out>, SN=288, svals=0x7ffffffed4e0, calreg_dims=<optimized out>, data=0x68024f0, imgcov=<optimized out>) at /home/simon/bart-0.4.03/src/calib/calib.c:318 #9 calib2 (conf=<optimized out>, out_dims=<optimized out>, out_data=<optimized out>, eptr=0x7fffff650000, SN=<optimized out>, svals=<optimized out>, calreg_dims=<optimized out>, data=<optimized out>, msk_dims=<optimized out>, msk=<optimized out>) at /home/simon/bart-0.4.03/src/calib/calib.c:537 #10 0x0000000000445e5e in calib (conf=0x2, out_dims=0x7ffffffec910, out_data=0x0, eptr=0x8, SN=0, svals=0x7ffffffec910, calreg_dims=0x7ffffffed960, data=0x6aa1f0) at /home/simon/bart-0.4.03/src/calib/calib.c:577 #11 0x0000000000420387 in main_ecalib (argc=<optimized out>, argv=<optimized out>) at src/ecalib.c:223 #12 0x000000000040f2a3 in main_bart (argc=<optimized out>, argv=<optimized out>) One more try with a 3D data set of dimension 308x384x64x16 (FE x PE x SE x Ch) natively in WSL, the core dumped occurs. simon@DESKTOP-6P96G2J:~/bart-0.4.03$ /home/simon/bart-0.4.03/bart ecalib -c0 -m1 /mnt/c/Users/simon/AppData/Local/Temp/tpc0842209_212e_4435_a59e_90081898a371in1 /mnt/c/Users/simon/AppData/Local/Temp/tpc0842209_212e_4435_a59e_90081898a371out1 /mnt/c/Users/simon/AppData/Local/Temp/tpc0842209_212e_4435_a59e_90081898a371out2 bart: /home/simon/bart-0.4.03/src/num/casorati.c:40: void casorati_dims(unsigned int, long *, const long *, const long *): Assertion `dim[i] >= dimk[i]' failed. Aborted (core dumped) gdb bt gives the following return: Thread 1 "bart" received signal SIGABRT, Aborted. __GI_raise (sig=sig@entry=6) at ../sysdeps/unix/sysv/linux/raise.c:51 51 ../sysdeps/unix/sysv/linux/raise.c: No such file or directory. (gdb) bt #0 __GI_raise (sig=sig@entry=6) at ../sysdeps/unix/sysv/linux/raise.c:51 #1 0x00007ffffdc7442a in __GI_abort () at abort.c:89 #2 0x00007ffffdc6be67 in __assert_fail_base (fmt=<optimized out>, assertion=assertion@entry=0x48d5ef "dim[i] >= dimk[i]", file=file@entry=0x48d601 "/home/simon/bart-0.4.03/src/num/casorati.c", line=line@entry=40, function=function@entry=0x48d62c "void casorati_dims(unsigned int, long *, const long *, const long *)") at assert.c:92 #3 0x00007ffffdc6bf12 in __GI___assert_fail (assertion=0x48d5ef "dim[i] >= dimk[i]", file=0x48d601 "/home/simon/bart-0.4.03/src/num/casorati.c", line=40, function=0x48d62c "void casorati_dims(unsigned int, long *, const long *, const long *)") at assert.c:101 #4 0x000000000045584c in casorati_dims (N=<optimized out>, odim=<optimized out>, dimk=<optimized out>, dim=<optimized out>) at /home/simon/bart-0.4.03/src/num/casorati.c:40 #5 0x0000000000448426 in calibration_matrix (calmat_dims=0x7ffffffe9490, kdims=<optimized out>, calreg_dims=0x7ffffffed910, data=0x6a3e50) at /home/simon/bart-0.4.03/src/calib/calmat.c:34 #6 covariance_function (kdims=<optimized out>, N=<optimized out>, cov=<optimized out>, calreg_dims=0x7ffffffed910, data=0x6a3e50) at /home/simon/bart-0.4.03/src/calib/calmat.c:161 #7 0x000000000044365e in compute_kernels (conf=0x7ffffffedd20, nskerns_dims=<optimized out>, nskerns_ptr=<optimized out>, SN=3456, val=0x7ffffffea310, caldims=<optimized out>, caldata=<optimized out>) at /home/simon/bart-0.4.03/src/calib/calib.c:693 #8 0x00000000004452e1 in calone (conf=0x7ffffffedd20, cov_dims=<optimized out>, SN=3456, svals=0x7ffffffea310, calreg_dims=<optimized out>, data=0x7ffffffe9e60, imgcov=<optimized out>) at /home/simon/bart-0.4.03/src/calib/calib.c:318 #9 calib2 (conf=<optimized out>, out_dims=<optimized out>, out_data=<optimized out>, eptr=0x7fff67990000, SN=<optimized out>, svals=<optimized out>, calreg_dims=<optimized out>, data=<optimized out>, msk_dims=<optimized out>, msk=<optimized out>) at /home/simon/bart-0.4.03/src/calib/calib.c:537 #10 0x0000000000445e5e in calib (conf=0x2, out_dims=0x7ffffffe9140, out_data=0x0, eptr=0x8, SN=0, svals=0x7ffffffe9140, calreg_dims=0x7ffffffed910, data=0x6a3e50) at /home/simon/bart-0.4.03/src/calib/calib.c:577 #11 0x0000000000420387 in main_ecalib (argc=<optimized out>, argv=<optimized out>) at src/ecalib.c:223 #12 0x000000000040f2a3 in main_bart (argc=<optimized out>, argv=<optimized out>) Thanks. Sent from Mail<https://go.microsoft.com/fwlink/?LinkId=550986> for Windows 10

…

________________________________ From: Martin Uecker <[email protected]> Sent: Wednesday, August 8, 2018 5:09:57 AM To: mrirecon/bart Cc: smyeungx; Author Subject: Re: [mrirecon/bart] Segmentation fault on WSL (Debian and Ubuntu) (#140) Let me know whether this worked. Then we could document this as a workaround. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#140 (comment)>, or mute the thread<https://github.com/notifications/unsubscribe-auth/AoLSov6d6wX16pj5Slv3xElnBayvJFKwks5uOgIlgaJpZM4VvthK>.

[johannestoger]

Hi, I am also hitting this bug. Just wanted to +1 it. Please let me know if there is any experiments I can do to help out.

[uecker]

Hi, to work around the segfault issue when compiling BART on WSL with GCC due to a deficiency of WSL (it does not support executable stacks) and if clang doesn't work, you can try using an experimental NOEXEC_STACK feature (on the noexecstack branch):

git fetch
git checkout noexecstack
make allcean
CFLAGS=-DNOEXEC_STACK make

[johannestoger]

I can add that compling with clang worked much better.