Read unsigned integers in WKBReader

[dbaston]

OGC WKB spec specifies that the integers are unsigned. Use of signed
integers is probably a carryover from Java, which does not have unsigned
integers. Clear type conversion warnings in WKBReader and use C++11
int64_t type.

[pramsey]

I guess the fix for OOM is still to come? It looks like CoordinateSequence reading

https://github.com/dbaston/libgeos/blob/wkbreader-unsigned/src/io/WKBReader.cpp#L431

starts by allocating a new CoordinateArraySequence with a (size, dims) constructor

https://github.com/dbaston/libgeos/blob/wkbreader-unsigned/src/geom/CoordinateArraySequence.cpp#L36-L41

which initializes the underlying vector using the (count) constructor

https://en.cppreference.com/w/cpp/container/vector/vector

Which does

"4) Constructs the container with count default-inserted instances of T. No copies are made."

So if you feed a degenerate WKB linestring with a 2B count, you can get GEOS to allocate the whole 2B before you actually try and read them and find the end on the input stream.

[dr-jts]

I guess the fix for OOM is still to come?

JTS has fixed this in PR 675, at least for the case when the size of the input byte sequence is known. It computes an upper bound on the value of all num* fields in the WKB, based on the size of the input. The num* fields are then checked against that limit.

[dbaston]

I guess the fix for OOM is still to come?

Right, this is unrelated to OOM.