summary hmr_count is non-zero but hmr output file is empty

[cb4github]

Describe the bug
See subject.

To Reproduce
Steps to reproduce the behavior (steps 1-8 for data preparation):

1. wget --directory-prefix=data https://ftp.ncbi.nlm.nih.gov/geo/samples/GSM5652nnn/GSM5652231/suppl/GSM5652231%5FCerebellum%2DNeuron%2DZ000000TB.hg38.beta
2. wgbstools init_genome --fasta_path $genome_reference_raw_fasta_file_path --force --no_default --debug --threads 19 hg38
3. ln -s $wgbs_tools_hg38_references_path $wgbs_tools_default_references_path
4. wgbstools beta2bed --outpath $methylation_counts_file_path --genome hg38 $beta_file_path
5. wgbstools beta2bed --mean --outpath $methylation_mean_file_path --genome hg38 $beta_file_path
6. paste
   <(gawk 'OFS="\t" {print $1,$2,"+", "CpG",$4}' $methylation_mean_file_path)
   <(cut -f 5 $methylation_counts_file_path) \

   $methylation_mean_plus_coverage_file_path

7. head output/GSM5652231_Cerebellum-Neuron-Z000000TB.hg38.methylation_mean_plus_coverage.meth
   chr1 10468 + CpG 0.5 16
   chr1 10470 + CpG 0.857 14
   chr1 10483 + CpG 0.714 14
   chr1 10488 + CpG 0.786 14
   chr1 10492 + CpG 0.714 14
   chr1 10496 + CpG 1 15
   chr1 10524 + CpG 1 15
   chr1 10541 + CpG 0.923 13
   chr1 10562 + CpG 0.562 16
   chr1 10570 + CpG 0.765 17
8. wc output/GSM5652231_Cerebellum-Neuron-Z000000TB.hg38.methylation_mean_plus_coverage.meth
   856 5136 20094 output/GSM5652231_Cerebellum-Neuron-Z000000TB.hg38.methylation_mean_plus_coverage.meth
9. singularity exec $DNMTOOLS_IMAGE dnmtools hmr
   -p ${output_directory}/params.txt
   -o $hmr_file_path
   -verbose
   -summary ${output_directory}/summary.txt
   $methylation_mean_plus_coverage_file_path
10. cat output/summary.txt
    hmr_count: 4
    hmr_total_size: 1713
    hmr_mean_size: 428.25
11. ls -s output/GSM5652231_Cerebellum-Neuron-Z000000TB.hg38.hmr
    0 output/GSM5652231_Cerebellum-Neuron-Z000000TB.hg38.hmr

Expected behavior
Resulting hmr file non-empty.

Screenshots
N/A

Desktop (please complete the following information):

• OS: centos7
• module show dnmtools/1.4.2

---

/lustre/project/PI/apps/centos7/modulefiles/dnmtools/1.4.2:

module-whatis DNMTools is a set of tools for analyzing DNA methylation data from high-throughput sequencing experiments, especially whole genome bisulfite sequencing (WGBS), but also reduced representation bisulfite sequencing (RRBS).
module load anaconda3/2023.07 singularity/3.9.0
prepend-path PATH /lustre/project/PI/apps/dnmtools/1.4.2/bin
prepend-path PATH /lustre/project/PI/apps/dnmtools/1.4.2/bin/utils
setenv DNMTOOLS_IMAGE /lustre/project/PI/singularity_images/dnmtools_latest_plus_bash.sif

• cat /lustre/project/PI/singularity_images/dnmtools_latest_plus_bash.def
  Bootstrap: docker
  From: ghcr.io/smithlabcode/dnmtools:latest

%post
apk update
apk add bash
apk add bash-doc
apk add bash-completion

%test
bash

Smartphone (please complete the following information):

• N/A

Additional context
Please let me know if you need any more information. Thanks.

Best,
CB

[andrewdavidsmith]

@cb4github Thanks for this! I know the source of the problem and will prioritize fixing it.

[andrewdavidsmith]

Fixed with commit #214 and a release will happen soon to propagate this fix. Thank again @cb4github this really helps!

[cb4github]

@andrewdavidsmith Many thanks for fixing this. I'm currently unable to install from the resulting master without an updated tar ball or Dockerfile or working configure script. Thanks.
Best,
CB

[cb4github]

Update: I was able to reproduce issue #209 using gcc/8.5.0, which supports C++17 as mentioned in the README.

Yet, as I understand it, this appears to be inconsistent with the very-cheap optimization not being available until gcc 12.1.

Here, I believe, is a relevant quote from the RH developers' website here.

Only relatively recently in GCC 12.1 was the auto-vectorizer enabled when -O2 is specified.

Here is sample output from that attempted make.

make[4]: Entering directory `/lustre/project/<PI>/build/centos7/dnmtools/src/abismal'
depbase=`echo src/abismal.o | sed 's|[^/]*$|.deps/&|;s|\.o$||'`;\
        g++ -std=c++17 -DHAVE_CONFIG_H -I.  -I ./src/smithlab_cpp -I ./src/bamxx  -fopenmp -O3  -MT src/abismal.o -MD -MP -MF $depbase.Tpo -c -o src/abismal.o src/abismal.cpp &&\
        mv -f $depbase.Tpo $depbase.Po
In file included from src/abismal.cpp:34:
src/AbismalAlign.hpp:215:50: error: unknown vectorizer cost model 'very-cheap'
 #pragma GCC optimize("vect-cost-model=very-cheap")
                                                  ^
src/AbismalAlign.hpp:215:50: note: valid arguments to '-fvect-cost-model=' are: cheap dynamic unlimited; did you mean 'cheap'?

Please advise. Thanks.

[andrewdavidsmith]

@cb4github Would you be able to move that to the #209 or open a new issue? I've got a reply, but want to keep things separate if that's ok.