From: Stephan T. Lavavej (stl_at_[hidden])
Date: 2006-10-08 20:30:58
> 1570 failures in 26 libraries (1519 are from non-broken
> Now, this looks scary, onless one figures out that there is
> an error in interpreting regression results, and only 51
> failures seem to be from non-broken platforms :-)
> The wording is intended to scare developers into fixing bugs
Arkadiy's point is not that the 1570 looks scary, but that it should say
"1570 failures in 26 libraries (51 are from non-broken platforms)"
"1570 failures in 26 libraries (1519 are from broken platforms)"
The summary incorrectly says that the broken-number is the
non-broken-number. This error is not repeated in the per-library results
("serialization (8 of 986 failures are from non-broken platforms)").
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk