John Naylor <jcnaylor@gmail.com> writes:
> On 4/26/18, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> I notice that duplicate_oids is now a good factor of 10 slower than it was
>> before (~0.04 sec to ~0.4 sec, on my machine). While this doesn't seem
>> like a problem for manual use, it seems annoying as part of the standard
>> build process, especially on slower buildfarm critters.
> If you're concerned about build speed, I think the generated *_d.h
> headers cause a lot more files to be rebuilt than before. I'll think
> about how to recover that.
Personally, I use ccache which doesn't seem to care too much, but I agree
than in some usages, extra touches of headers would be costly. Perhaps
it's worth adding logic to avoid overwriting an existing output file
unless it changed? I'm not sure; that would cost something too.
> -removes duplicate_oids (not sure if you intended this, but since
> unused_oids has already been committed to report duplicates, we may as
> well standardize on that) and cleans up after that fact
I don't think anyone's asked for duplicate_oids to be removed altogether,
and I'm afraid that doing so would break existing workflows. For that
matter, now that I think about it, changing the behavior of unused_oids
as we did yesterday was ill-considered as well, so I undid that part.
I've pushed the current-directory-independence change by itself so that
we can see if any buildfarm members think it's problematic. (They'll
still be testing duplicate_oids, as things stand.) Once we have some
confirmation on the FindBin stuff actually being portable, I'll push
the changes to make genbki do duplicate checks and remove duplicate_oids
from the build sequence.
regards, tom lane