Upgrade shapelib

[strk]

New code is better :) See ​https://github.com/OSGeo/shapelib/pull/3 which was contributed by a PostGIS user initially as a pull request for our copy: ​https://git.osgeo.org/gitea/postgis/postgis/pulls/40

[pramsey]

I just tried to copy in upstream, but it broke us :(

 ./shp2pgsql ../regress/loader/Point
Shapefile type: Point
Postgis type: POINT[2]
SET CLIENT_ENCODING TO UTF8;
SET STANDARD_CONFORMING_STRINGS TO ON;
BEGIN;
CREATE TABLE "." (gid serial);
ALTER TABLE "." ADD PRIMARY KEY (gid);
SELECT AddGeometryColumn('','.','geom','0','POINT',2);
COMMIT;
ANALYZE ".";

Almost like it's not bothering to iterate on the file.

[pramsey]

Just checked in debugger, it's iterating, but it thinks every record is deleted due to a DBF file marker.

[Paul Ramsey <pramsey@…>]

In 00eb4b4/git:

Update the shapelib copy to the latest.
References #4591.
Required two changes to pass regression.

• Modify dbfopen.c so that DBFIsRecordDeleted() does not test iShape >= psDBF->nRecords unless psDBF->nRecords > 0.
• Add DBFSetWriteEndOfFileChar(state->dbf, 0) to pgsql2shp-core.c after dbf opening, since apparently we did not write the EOF marker in earlier versions.

[Paul Ramsey <pramsey@…>]

In 42dd04bb/git:

Fix CI failure in SADFWrite,
References #4591

[Algunenano]

I think this broke Windows build: ​https://trac.osgeo.org/postgis/ticket/4603

[robe]

We didn't change shapelib yet in 3.0. Given the invasiveness of it, I say we keep it at 3.1 and don't backport it.

[pramsey]

I think this is done, new bugs can get new tickets.