Sometimes BLAST+ warning messages treated as errors

[peterjc]

I've just hit the following problem with BLAST+ 2.2.29 writing these warning messages to stderr which (sadly) start with Error: and so match our regex:

Error: (1431.1) FASTA-Reader: Warning: FASTA-Reader: Ignoring FASTA modifier(s) found because the input was not expected to have any.

Currently in ncbi_macros.xml we have:

    <stdio>
        <!-- Anything other than zero is an error -->
        <exit_code range="1:" />
        <exit_code range=":-1" />
        <!-- In case the return code has not been set propery check stderr too -->
        <regex match="Error:" />
        <regex match="Exception:" />
    </stdio>

I will report this to the NCBI, see http://blastedbio.blogspot.co.uk/2014/04/error-1431-fasta-reader-ignoring-modifiers.html for details. However, perhaps as a short term workaround the Error: regex could be relaxed if Warning: also appears in the same line?

[peterjc]

This problem was fixed in BLAST+ 2.2.30, so I am closing this issue for now.

[peterjc]

Reopening: Similar issue reported on galaxy-dev list http://dev.list.galaxyproject.org/Cosmetic-Makeblastdb-error-promulgated-through-wrapper-tc4667485.html by Damion Dooley (@ddooley), and blogged about by Keith Bradnam (@kbradnam) here: http://www.acgt.me/blog/2014/5/15/fun-with-an-error-message-from-ncbi-blast

Error: (1431.1) FASTA-Reader: Warning: FASTA-Reader: First data line in seq is about 45% ambiguous nucleotides (shouldn't be over 40%)

Again, this looks like a warning which the NCBI BLAST suite is wrongly labelling as an error.

[kbradnam]

I emailed the NCBI about this on 2014-05-14 and had the following reply:

Thank you for the feedback. I agree that this message should have more of a warning tone. Will discuss with the developers.

[peterjc]

Thanks Keith. Hopefully they'll fix this in the next BLAST+ release, but I'm open to adjusting our regex so these Error: ... Warning: ... cases are not treated as errors by the Galaxy wrapper.

[Public-Health-Bioinformatics]

As noted on galaxy board,

<xml name="stdio">
    <stdio>
        <!-- Anything other than zero is an error -->
        <exit_code range="1:" />
        <!-- Might see negative return codes for Unix signals via Python subprocess -->
        <exit_code range=":-1" />
        <!-- In case the return code has not been set properly check stderr too -->
    <regex match="^Error: \(1431.1\) FASTA-Reader: Warning: FASTA-Reader: First data line in seq is about" level="log" />
        <regex match="^Error:" />
        <regex match="Exception:" />
    </stdio>
</xml>

provides a workaround until NCBI downgrades the Error:.

The regex line has to be constrained with a carrot because otherwise it matches the too-generic first line of error report.

I can set myself up to do a pull request on this if desired. Will have to wait a few weeks as I have other things on the go though, incl. conference and holiday.

ciao,

Damion

[peterjc]

Thanks Damion - I'm also on the road (currently at the Galaxy conference GCC2015 in Norwich), but adjusting the regex should be fine. Thanks!