Re: Inconsistencies around defining FRONTEND

Hi,

On 2022-08-20 12:45:50 -0700, Andres Freund wrote:
> The -DFRONTENDs for the various ecpg libraries don't seem necessary
> anymore. That looks to be a leftover from 7143b3e8213, before that ecpg had
> copies of various pgport libraries.
>
> Same with libpq, also looks to be obsoleted by 7143b3e8213.
>
> I don't think we need FRONTEND in initdb - looks like that stopped being
> required with af1a949109d.

I think the patches for this are fairly obvious, and survived CI without an
issue [1], so the src/tools/msvc bits work too. So I'm planning to push them
fairly soon.

But the remaining "issues" don't have an obvious solutions and not addressed
by these patches:

> Unfortunately, the remaining uses of FRONTEND are required. That's:
> - pg_controldata, via #define
> - pg_resetwal, via #define
> - pg_rewind, via -DFRONTEND, due to xlogreader.c
> - pg_waldump, via #define and -DFRONTEND, due to xlogreader.c, xlogstats.c, rmgrdesc/*desc.c
>
> I'm kind of wondering if we should add xlogreader.c, xlogstat.c, *desc to
> fe_utils, instead of building them in various places. That'd leave us only
> with #define FRONTENDs for cases that do need to include postgres.h
> themselves, which seems a lot more palatable.

Greetings,

Andres Freund

[1] https://cirrus-ci.com/build/4648937721167872