search

ashander / ftprime

Forward-time simulation of the msprime data structure (for development)
2 stars 1 forks source link

Example of test failure output; occurs 1/10 times or so #12

Open ashander opened 7 years ago

ashander commented 7 years ago

@petrelharp two things:

1) I added a quick commit updating the version of msprime that your latest changes depend on to current master 2150a014094f026e9e19c01809d2df67b3e57190 See git submodule for details 2) assuming that's the right version, should tests always pass? I get occasional failures. See the file in this PR

petrelharp commented 7 years ago

Oh, that's great. I was encountering such errors in the much larger sims produced by simupop and was despairing ever finding what the problem was in such large ARGs. I suspect the error is in msprime, but we'll see. I'll debug; but if you need this short-term you can fall back to earlier versions.

petrelharp commented 7 years ago

Oh, right. The problem is that msprime determines the number of samples automatically as the smallest-numbered internal node. I was not ensuring that this was the case.

This will change soon in msprime, so I am hacking something in.

petrelharp commented 7 years ago

I fixed that; there remain some issues at least with .simplify(): see https://github.com/jeromekelleher/msprime/pull/131