[postgis-devel] 3.0.1 issues report

Greg Troxel gdt at lexort.com
Fri Jun 26 08:52:13 PDT 2020


rmrodriguez at carto.com writes:

> Please let me know when you've had the chance of testing it. If it
> works correctly I'll push the fixes to 3.0 and the future 3.1.

I found the github web interface remarkably difficult in that there was
no obvious way to download a patch for those commits.  But I added your
repo as a remote to my clone of the official repo, and ran git diff....

With that applied to the 3.0.1 release, and rerunning autoconf, I was
able to build without pkg-config, json-c, protobuf-c exposd to the
build, and 'make check' passed.

>> It might be reasonable to just say 'pkg-config is required', test for
>> it, and error out if not.
>
> I think we still officially support dependencies that didn't use
> pkg-config (and they they started using it), but I think it would
> indeed make sense to start requiring pkg-config to simplify both the
> management and errors in different environments.

I didn't mean to require that all dependencies be found via pkg-config.
I meant to require pkg-config to be present to build postgis at all.
Then each dependency can be moved to pkg-config only as that is
reasonable.

>> I guess it would be nice if README.postgis explained really clearly what
>> was gained or lost by having it.  (I realize I tend to be particularly
>> demanding about these sorts of doc/advice things.)
>
> I see that README.postgis does mention that protobuf-c is necessary
> for ST_AsMVT and ST_AsGeoBuf. How would you reword it to make it
> clearer?

I guess my request is unreasonable, but I'd like someone who is building
binaries for others to be able to tell if they should

  1) impose the protobuf dependency on everyone, letting those two
  functions work, or

  2) omit protobuf, have those functions not work, and have everybody
  not have the dependency

which is really a question of whether being handed a binary postgis
install that is missing protobuf is 1) normal or 2) defective.  I'm
trying to separate "SHOULD have this dependency (but we can build
without it)" from "this is weird and if you want to do this, add it".

But really you can just tell me wich is which.


More information about the postgis-devel mailing list