[Debian-med-packaging] Bug#969608: makeblastdb 2.10.x on 32-bit architectures

[Aaron M. Ucko]

Étienne Mollier <etienne.mollier at mailoo.org> writes:

> I worked within an arm64 schroot built with sbuild-createchroot,
> on a native amd64 machine.

Thanks for clarifying.  AFAICT, this environment imposes a tighter limit
than native arm64 hardware, and versions 2.10.0-1 and 2.10.0-3 both hit
it.  Rough bisection via the BLASTDB_LMDB_MAP_SIZE environment variable
gives an empirical limit of 20,073,607,168 bytes (4,900,783 4K pages).
This number isn't particularly round, so it presumably reflects what
remains of some cumulative limit.  As such, the default should probably
be at most 20,000,000,000 bytes (4,882,812½ pages ;-) to build in more
of a margin.  That's 1/15 upstream's default, but with any luck should
be plenty in practice, so I'm open to making that adjustment.  Also,
this reduced limit would still be well more than we (can) allow on
32-bit architectures, which is in turn much more than upstream's trunk
allows on Windows:

https://www.ncbi.nlm.nih.gov/IEB/ToolBox/CPP_DOC/lxr/source/include/objtools/blast/seqdb_writer/writedb_lmdb.hpp#L51

-- 
Aaron M. Ucko, KB1CJC (amu at alum.mit.edu, ucko at debian.org)
http://www.mit.edu/~amu/ | http://stuff.mit.edu/cgi/finger/?amu@monk.mit.edu