Removal of HeuristicOverlay changing error messages downstream

[pramsey]

This PostGIS regression test shows a new error message after a fairly unrelated commit for the floating error handler.

https://trac.osgeo.org/postgis/ticket/5312

Also note in the old path, the exception was caught and validity checked to add to the info in the error message. This new error path seems to omit that, either accidentally or deliberately due to other changes? (I think this happened in the "heuristic overlay" code pasta, which was slated for reeducation?)

[dbaston]

I would guess that it is a result of #788, not #791. Is the change in error message a problem?

[pramsey]

Yes, you're right that's where it came from. One thing lost was the check for validity when exceptions are caught... https://github.com/libgeos/geos/pull/788/files#diff-fe4091b0d1cbe6972cabc0f5daba8f737ba035a9bb569055a16a3700987b9565L432

I'm not sure we (postgis) should be quite so concerned with error messages in general, though the existence of that test implies it exists all the way back into a lot of earlier versions, which all will need to be fixed up when this version drops, alas.

cc / @robe2

[dbaston]

which all will need to be fixed up when this version drops

I thought those tests were essentially doing assert message.startsWith(expected), so you could truncate the expected values to be "GEOS Error:" or "GEOS Error: TopologyException" and live in version-independent bliss. Maybe I'm misremembering, though.

[robe2]

I'm fine with truncating the messages in postgis. Did it for some others, but I'm not seeing the error anymore in PostGIS. @pramsey did you change something? debbie is all green - https://trac.osgeo.org/postgis/ticket/5312 so I guess we could close this out or more changes need to be done?

[pramsey]

Yes, I made the test less strict in 07d29463a26ecc47dc1fd43897d05493a3645700