baker-laboratory / RoseTTAFold-All-Atom

Other
614 stars 108 forks source link

psipred buffer overflow detected #72

Open cdienem opened 5 months ago

cdienem commented 5 months ago

Dear RoseTTAFold team,

when running the protein example after fixing issues #32 and #37 , I run into an error with psipred:

Running PSIPRED buffer overflow detected : psipred terminated ======= Backtrace: ========= /lib64/libc.so.6(fortify_fail+0x37)[0x7f282a2d4577] /lib64/libc.so.6(+0x1166f2)[0x7f282a2d26f2] /lib64/libc.so.6(__fgets_chk+0x129)[0x7f282a2d2a19] psipred[0x401390] psipred[0x40158a] /lib64/libc.so.6(libc_start_main+0xf5)[0x7f282a1de555] psipred[0x400a89] ======= Memory map: ======== 00400000-00402000 r-xp 00000000 fd:02 23127977 /home/X/opt_test/conda_envs/RFAA/bin/psipred 00601000-00602000 r--p 00001000 fd:02 23127977 /home/X/opt_test/conda_envs/RFAA/bin/psipred 00602000-00603000 rw-p 00002000 fd:02 23127977 /home/X/opt_test/conda_envs/RFAA/bin/psipred 00603000-006d3000 rw-p 00000000 00:00 0 02265000-02286000 rw-p 00000000 00:00 0 [heap] 7f282a1bc000-7f282a37f000 r-xp 00000000 fd:00 33588779 /usr/lib64/libc-2.17.so 7f282a37f000-7f282a57f000 ---p 001c3000 fd:00 33588779 /usr/lib64/libc-2.17.so 7f282a57f000-7f282a583000 r--p 001c3000 fd:00 33588779 /usr/lib64/libc-2.17.so 7f282a583000-7f282a585000 rw-p 001c7000 fd:00 33588779 /usr/lib64/libc-2.17.so 7f282a585000-7f282a58a000 rw-p 00000000 00:00 0 7f282a58a000-7f282a68b000 r-xp 00000000 fd:00 33588787 /usr/lib64/libm-2.17.so 7f282a68b000-7f282a88a000 ---p 00101000 fd:00 33588787 /usr/lib64/libm-2.17.so 7f282a88a000-7f282a88b000 r--p 00100000 fd:00 33588787 /usr/lib64/libm-2.17.so 7f282a88b000-7f282a88c000 rw-p 00101000 fd:00 33588787 /usr/lib64/libm-2.17.so 7f282a88c000-7f282a8ae000 r-xp 00000000 fd:00 33588491 /usr/lib64/ld-2.17.so 7f282aa8a000-7f282aa8d000 rw-p 00000000 00:00 0 7f282aa8f000-7f282aa93000 r--p 00000000 fd:02 541974413 /home/X/opt_test/conda_envs/RFAA/lib/libgcc_s.so.1 7f282aa93000-7f282aaa5000 r-xp 00004000 fd:02 541974413 /home/X/opt_test/conda_envs/RFAA/lib/libgcc_s.so.1 7f282aaa5000-7f282aaa8000 r--p 00016000 fd:02 541974413 /home/X/opt_test/conda_envs/RFAA/lib/libgcc_s.so.1 7f282aaa8000-7f282aaa9000 r--p 00019000 fd:02 541974413 /home/adminc/opt_test/conda_envs/RFAA/lib/libgcc_s.so.1 7f282aaa9000-7f282aaaa000 rw-p 0001a000 fd:02 541974413 /home/X/opt_test/conda_envs/RFAA/lib/libgcc_s.so.1 7f282aaaa000-7f282aaad000 rw-p 00000000 00:00 0 7f282aaad000-7f282aaae000 r--p 00021000 fd:00 33588491 /usr/lib64/ld-2.17.so 7f282aaae000-7f282aaaf000 rw-p 00022000 fd:00 33588491 /usr/lib64/ld-2.17.so 7f282aaaf000-7f282aab0000 rw-p 00000000 00:00 0 7fff68f21000-7fff68f42000 rw-p 00000000 00:00 0 [stack] 7fff68ffe000-7fff69000000 r-xp 00000000 00:00 0 [vdso] ffffffffff600000-ffffffffff601000 r-xp 00000000 00:00 0 [vsyscall]

The system is CentOS7. Does psipred come pre-compiled? If its compiled during installation of RoseTTAFOld-all-atom, the local gcc verison is 4.8.5.

Please let me know if you need more information.

felixmb1 commented 5 months ago

I get the same issue and have tried fixes from previous rosettafold release https://github.com/Sabryr/ProteinFolding/issues/3 but it still doesn't work

cdienem commented 5 months ago

Alright, I went through many issues so I will summarize what has worked for me to get rid of all of them.

The main issues are:

After fixing these, I successfully ran the protein example.

How to fix the above? First, install RFAA as described. Then apply the following fixes:

# activate your conda environment
conda activate RFAA

# Define where RFAA is installed
RFAA=path/to/rosettafold-all-atom
cd $RFAA

# The script input_prep/make_ss.sh needs to be executable
chmod +x input_prep/make_ss.sh

# The UniRef version in make_msa.sh is wrong. Rename to the correct UniRef version
sed -i "s|UniRef30_2021_06|UniRef30_2020_06|g" make_msa.sh

# The uniref database is expected to be "uniclust" and not "UniRef..." so symlink it
ln -s UniRef30_2020_06 uniclust

# Get new blast version as the one shipped with the installation does not work and make it available in input_prep/makse_ss.sh
wget https://ftp.ncbi.nlm.nih.gov/blast/executables/legacy.NOTSUPPORTED/2.2.26/blast-2.2.26-x64-linux.tar.gz
mkdir -p blast-2.2.26
tar -xf blast-2.2.26-x64-linux.tar.gz -C blast-2.2.26
cp -r blast-2.2.26/blast-2.2.26/ blast-2.2.26_bk
rm -r blast-2.2.26
mv blast-2.2.26_bk/ blast-2.2.26

# Insert sourcing of the correct blast (pay attention to the location delivered by $RFAA)
new_line="export BLASTMAT=$RFAA/blast-2.2.26/data/"
sed -i "3 a$new_line" $RFAA/input_prep/make_ss.sh

# Recompile psipred in case the shipped one runs into a buffer overflow (I guess it would be better to always compile during installation)
cd $RFAA
wget http://bioinfadmin.cs.ucl.ac.uk/downloads/psipred/psipred.4.02.tar.gz
tar xf psipred.4.02.tar.gz
cd psipred/src
make
make install

chmod +x psipred
chmod +x psipass2

# Replace the psipred binaries within the conda environment with the recompiled ones (symlink them)

mv $CONDA_PREFIX/bin/psipred $CONDA_PREFIX/bin/psipred.orig
mv $CONDA_PREFIX/bin/psipass2 $CONDA_PREFIX/bin/psipass2.orig

ln -s $RFAA/psipred/bin/psipred $CONDA_PREFIX/bin/psipred
ln -s $RFAA/psipred/bin/psipass2 $CONDA_PREFIX/bin/psipass2

So this should be easy to implement now for the devs.

NiklasTR commented 5 months ago

Took me steps 1, 2, and 3 to find this issue - great work. Worth a PR