Hi Mark,

thank you for the review!

>> +     `(;; There are three(!) tests for this massive library, and all fail with
>> +       ;; "unparsable timing stats".
>> +       ;; ERR [127] --  [util/regexp] test_pcre.sh     (unparsable timing stats)
>> +       ;; ERR [127] --  [serial/datatool] datatool.sh     (unparsable timing stats)
>> +       ;; ERR [127] --  [serial/datatool] datatool_xml.sh     (unparsable timing stats)
>> +       #:tests? #f
>
> Just a guess, but maybe this is because you replaced "/bin/date" with
> "echo -n 0".  How about replacing it with "date -d @0" instead?

I tried that but I still get the same problem.  The test script is
generated from a template in common/check/check_make_unix.sh.  We are
substituting "/var/tmp" for "/tmp" and the "(/usr)/bin" prefix, but this
should not have an impact on the functionality of the generated script.

I'm not sure why it's failing -- it works fine in "guix environment".
It's very time-consuming to recompile the whole thing until the tests
are reached.

> It would be great to get the tests working, even if we have to disable
> some of them.  Otherwise we have no way of knowing that we're not
> distributing broken garbage :)

It's hard to know in this case even with the tests, because it's only
three tests and they seem hardly representative of the library.

>> +
>> +            ;; rewrite "/var/tmp" in check script
>> +            (substitute* "scripts/common/check/check_make_unix.sh"
>> +              (("/var/tmp") (string-append (getcwd) "/build/build")))
>
> Or maybe just "/tmp" ?

Yes, that also works.

> All of these plus the ones for 'sh' could be combined into something
> like this: (untested)

[...]

This works great!  Thank you.

>> +              (("^ *PATH=.*") "")
>> +              (("action=/bin/") "action=")
>> +              (("export PATH") "echo -n 0"))
>
> Why "echo -n 0" here?  Maybe ":" would be better?  It is a no-op
> built-in command in Bourne shell.

Yes, this works.  I needed to replace it with a no-op because "export
PATH" is also found in the middle of a long chain of commands, so
replacing it with the empty string results in a syntax error.

> Is everything in here really in the public domain?  I'd guess that in
> order to make this true, you'd need to remove bzip2 and zlib in a
> snippet, and even then I'd doubtful :)

I've moved the code to remove the bundled stuff to a snippet.

The NCBI code was released into the public domain.  However, it appears
that some third-party headers and some build scripts are under a
different license:

  * Expat:
    * ncbi-blast-2.2.30+-src/c++/include/util/bitset/
    * ncbi-blast-2.2.30+-src/c++/src/html/ncbi_menu*.js
  * Boost license:
    * ncbi-blast-2.2.30+-src/c++/include/util/impl/floating_point_comparison.hpp
  * LGPL 2+:
    * ncbi-blast-2.2.30+-src/c++/include/dbapi/driver/odbc/unix_odbc/
  * ASL 2.0:
    * ncbi-blast-2.2.30+-src/c++/src/corelib/teamcity_*

I could not find mention of any other licenses.  Is this correct, then:

    ;; Most of the sources are in the public domain, with the following exceptions:
    ;; ...(the above list)...
    (license (list license:public-domain
                   license:expat
                   license:boost1.0
                   license:lgpl2.0+
                   license:asl2.0))

What do you think?

~~ Ricardo