|From:||Heikki Linnakangas <hlinnaka(at)iki(dot)fi>|
|To:||Stephen Frost <sfrost(at)snowman(dot)net>, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>|
|Cc:||Robert Haas <robertmhaas(at)gmail(dot)com>, "pgsql-hackers(at)postgresql(dot)org" <pgsql-hackers(at)postgresql(dot)org>, Piotr Stefaniak <postgres(at)piotr-stefaniak(dot)me>|
|Subject:||Re: Preliminary results for proposed new pgindent implementation|
|Views:||Raw Message | Whole Thread | Download mbox | Resend email|
On 05/19/2017 06:05 PM, Stephen Frost wrote:
> * Tom Lane (tgl(at)sss(dot)pgh(dot)pa(dot)us) wrote:
>> Robert Haas <robertmhaas(at)gmail(dot)com> writes:
>>> On Thu, May 18, 2017 at 11:00 PM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>>>> The reason PGSSTrackLevel is "unrecognized" is that it's not in
>>>> typedefs.list, which is a deficiency in our typedef-collection
>>>> technology not in indent. (I believe the problem is that there
>>>> are no variables declared with that typename, causing there to
>>>> not be any of the kind of symbol table entries we are looking for.)
>>> This, however, doesn't sound so good. Isn't there some way this can be fixed?
>> I'm intending to look into it, but I think it's mostly independent of
>> whether we replace pgindent itself. The existing code has the same
>> problem, really.
>> One brute-force way we could deal with the problem is to have a "manual"
>> list of names to be treated as typedefs, in addition to whatever the
>> buildfarm produces. I see no other way than that to get, for instance,
>> simplehash.h's SH_TYPE to be formatted as a typedef. There are also
>> some typedefs that don't get formatted correctly because they are only
>> used for wonky options that no existing typedef-reporting buildfarm member
>> builds. Manual addition might be the path of least resistance there too.
>> Now the other side of this coin is that, by definition, such typedefs
>> are not getting used in a huge number of places. If we just had to
>> live with it, it might not be awful.
> Dealing with the typedef lists in general is a bit of a pain to get
> right, to make sure that new just-written code gets correctly indented.
> Perhaps we could find a way to incorporate the buildfarm typedef lists
> and a manual list and a locally generated/provided set in a simpler
> fashion in general.
You can get a pretty good typedefs list just by looking for the pattern
"} <type name>;". Something like this:
grep -o -h -I --perl-regexp -r "}\W(\w+);" src/ contrib/ | perl -pe
's/}\W(.*);/\1/' | sort | uniq > typedefs.list
It won't cover system headers and non-struct typedefs, but it catches
those simplehash typedefs and PGSSTrackLevel. Maybe we should run that
and merge the result with the typedef lists we collect in the buildfarm.
|Next Message||Tom Lane||2017-05-19 15:48:54||Re: Preliminary results for proposed new pgindent implementation|
|Previous Message||Tom Lane||2017-05-19 15:41:21||Re: PG10 Crash-safe and replicable Hash Indexes and UNIQUE|