daviesrob

Yes, once you've closed the `stdout` file descriptor, you can't reopen it again. What are you trying to do here? Trying to write two BCF files to stdout is unlikely...

I'm not sure exactly what checks you're planning to make, but if you're trying to test if files can be opened then it would be risky to assume that they...

Yes, it's very important to check all I/O functions. Other ones we need to pay more attention to are khash and kstring. With both of those, if malloc fails the...

You can get more debugging output by using the `htsfile` program that comes with htslib. This should get you a lot of output: ``` htsfile -vvvvvv -c s3://mybucket/my.cram ``` Htslib...

Which one does it fix, the `make test`, or `valgrind`? The problem with `make test` is a known one, caused by BCF's rather unfortunate (ab)use of signalling NANs (see samtools/hts-specs#145)....

It looks like `valgrind`has spotted a real problem. `write_format_values()` in `test-vcf-api.c` does this: ``` float test[4]; bcf_float_set_missing(test[0]); test[1] = 47.11f; bcf_float_set_vector_end(test[2]); bcf_update_format_float(hdr, rec, "TF", test, 4); bcf_write1(fp, hdr, rec); ```...

@pd3 The problem is the values that come _after_ the vector_end. Should they all be written as `vector_end` too, which is how they end up when read back in? Or...

No, mainly because it's not easy to fix. This signalling NAN issue arises due to reliance on behaviour which its not well-defined, especially on i386. Even if it's possible to...

I'm afraid the only way I've found to reliably fix this using compiler flags is to use: ``` ./configure CFLAGS='-g -O2 -msse -mfpmath=sse' ``` Packages that depend on htslib and...

Yes, the fact that `htslib-plugins` needs its own copy of `hfile_internal.h` suggests that something is missing for plug-in developers. Given that the interface isn't going to change, moving the bits...