[geos-devel] MSVC Unit Failure

[Paul Ramsey]

> On Dec 1, 2020, at 1:47 AM, Mike Taves <mwtoews at gmail.com> wrote:
> 
> This issue is very tricky. Running the command from a local Windows build:
>> test_geos_unit "geos::simplify::TopologyPreservingSimplifier"
> 
> will mostly fail (showing "tests summary: failures:1 ok:14"), but
> occasionally pass (showing "tests summary: ok:15"), with a pass/fail
> ratio close to 1:8. This type of random behaviour can only be
> explained by a memory error, but is only an issue with Windows.
> (Running the command on Linux with valgrind shows "All heap blocks
> were freed -- no leaks are possible")
> 
> Looking at the AppVeyor history is slightly revealing:
> https://ci.appveyor.com/project/dbaston/geos/history
> 
> The first captured instance of this issue was 4 months ago on the
> overlay-sr branch:
> https://ci.appveyor.com/project/dbaston/geos/builds/34227033/job/pq00xb0pirk5xpe9#L1032
> 
> It likely existed before this commit, up to (but not beyond) 6 months
> ago, but was obscured by a different preceding error. Nevertheless, it
> seems to have originated somewhere in the overlay-sr branch?

Well, that time period is when most of overlayng was ported, so there's a great deal of activity on that branch (which was the main porting branch).
Even doing a bisect could be hard because there will be times when the branch doesn't build. I don't now how anything other than a bisect will track this down. Unless there's a windows equivalent to valgrind that might find the memory error directly.

P

> _______________________________________________
> geos-devel mailing list
> geos-devel at lists.osgeo.org
> https://lists.osgeo.org/mailman/listinfo/geos-devel