On Sat, Mar 16, 2019 at 9:12 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:
> I wrote:
> > Good point. I also see jsonpath_gram.h left behind after maintainer-clean:
>
> Oh, and of potentially more significance: after maintainer-clean and
> re-configure, make fails with
>
> In file included from jsonpath_gram.y:24:
> ../../../../src/include/utils/jsonpath_scanner.h:25:33: error: utils/jsonpath_gram.h: No such file or directory
>
> I first thought this was a problem with insufficient dependencies
> allowing parallel make to do things in the wrong order, but the problem
> repeats even without any parallelism. It looks like the dependencies
> have been constructed in such a way that if the symlink at
> src/include/utils/jsonpath_gram.h exists but the underlying file
> does not, nothing will make the underlying file. This is pretty broken;
> aside from this outright failure, it also suggests that nothing will
> update that file if it exists but is out of date relative to its
> sources.
>
> Please make sure that the make rules associated with these files look
> exactly like the previously-debugged rules for existing bison/flex
> output files. There are generally good reasons for every last bit
> of weirdness in those.
I've pushed a fix. I hope I didn't forget anything.
BTW, it appears that windows build scripts also needs some fixup. I'm
not very familiar with that. I would be glad if somebody review the
attached patch.
------
Alexander Korotkov
Postgres Professional: http://www.postgrespro.com
The Russian Postgres Company
