Test suite failure on 32-bit machines (floating point differences) #445
Comments
|
Ah so that's what all the filtering was for. I did wonder given it appeared to work without it, although I was also hoping that we could have at least one full test that checked the entire file without having to filter lots away. Mutter! (Arguably it's a bug to output more bits of precision than are valid.) I guess in the interests of sanity we'll just have to remove testing of the variable fields. |
Yes. …sigh… |
jmarshall
added a commit
that referenced
this issue
Aug 3, 2015
charles-plessy
added a commit
to Debian/samtools
that referenced
this issue
Apr 26, 2016
Samtools release 1.3: many improvements, fixes, new commands * The obsolete "samtools sort in.bam out.prefix" usage has been removed. If you are still using -f, -o, or out.prefix, convert to use -T PREFIX and/or -o FILE instead. (samtools#295, samtools#349, samtools#356, samtools#418, PR samtools#441; see also discussions in samtools#171, samtools#213.) * The "bamshuf" command has been renamed to "collate" (hence the term bamshuf no longer appears in the documentation, though it still works on the command line for compatibility with existing scripts). * The mpileup command now outputs the unseen allele in VCF/BCF as <*> rather than X or <X> as previously, and now has AD, ADF, ADR, INFO/AD, INFO/ADF, INFO/ADR --output-tags annotations that largely supersede the existing DV, DP4, DPR annotations. * The mpileup command now applies BAQ calculations at all base positions, regardless of which -l or -r options are used (previously with -l it was not applied to the first few tens of bases of each chromosome, leading to different mpileup results with -l vs. -r; samtools#79, samtools#125, samtools#286, samtools#407). * Samtools now has a configure script which checks your build environment and facilitates choosing which HTSlib to build against. See INSTALL for details. * Samtools's Makefile now fully supports the standard convention of allowing CC/CPPFLAGS/CFLAGS/LDFLAGS/LIBS to be overridden as needed. Previously it listened to $(LDLIBS) instead; if you were overriding that, you should now override LIBS rather than LDLIBS. * A new addreplacerg command that adds or alters @rg headers and RG:Z record tags has been added. * The rmdup command no longer immediately aborts (previously it always aborted with "bam_get_library() not yet implemented"), but remains not recommended for most use (samtools#159, samtools#252, samtools#291, samtools#393). * Merging files with millions of headers now completes in a reasonable amount of time (samtools#337, samtools#373, samtools#419, samtools#453; thanks to Nathan Weeks, Chris Smowton, Martin Pollard, Rob Davies). * Samtools index's optional index output path argument works again (samtools#199). * Fixed calmd, targetcut, and potential mpileup segfaults when given broken alignments with POS far beyond the end of their reference sequences. * If you have source code using bam_md.c's bam_fillmd1_core(), bam_cap_mapQ(), or bam_prob_realn_core() functions, note that these now take an additional ref_len parameter. (The versions named without "_core" are unchanged.) * The tview command's colour scheme has been altered to be more suitable for users with colour blindness (samtools#457). * Samtools depad command now handles CIGAR N operators and accepts CRAM files (samtools#201, samtools#404). * Samtools stats now outputs separate "N" and "other" columns in the ACGT content per cycle section (samtools#376). * Added -a option to samtools depth to show all locations, including zero depth sites (samtools#374). * New samtools dict command, which creates a sequence dictionary (as used by Picard) from a FASTA reference file. * Samtools stats --target-regions option works again. * Added legacy API sam.h functions sam_index_load() and samfetch() providing bam_fetch()-style iteration over either BAM or CRAM files. (In general we recommend recoding against the htslib API directly, but this addition may help existing libbam-using programs to be CRAM-enabled easily.) * Fixed legacy API's samopen() to write headers only with "wh" when writing SAM files. Plain "w" suppresses headers for SAM file output, but this was broken in 1.2. * "samtools fixmate - -" works in pipelines again; with 1.0 to 1.2, this failed with "[bam_mating] cannot determine output format". * Restored previous "samtools calmd -u" behaviour of writing compression level 0 BAM files. Samtools 1.0 to 1.2 incorrectly wrote raw non-BGZF BAM files, which cannot be read by most other tools. (Samtools commands other than calmd were unaffected by this bug.) * Restored bam_nt16_nt4_table[] to legacy API header bam.h. * Fixed bugs samtools#269, samtools#305, samtools#320, samtools#328, samtools#346, samtools#353, samtools#365, samtools#392, samtools#410, samtools#445, samtools#462, samtools#475, and samtools#495.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
f845404 introduced a failure on 32-bit machines in the test case it changed, mpileup number 52. The expected output, presumably generated on a 64-bit machine, contains:
but on lucid-dev32 this position is calculated as VDB=0.768076.
VDB is one of the fields that was filtered by bcf_filter.sh and by
vcf-miniview -f… perhaps we need to be using-fhere after all.The text was updated successfully, but these errors were encountered: