Horton

Re: [sqlsmith] Planner crash on foreign table join

Andrew Gierth <andrew@tao11.riddles.org.uk>

View thread View on postgr.es
From: Andrew Gierth <andrew@tao11.riddles.org.uk>
To: Thomas Munro <thomas.munro@enterprisedb.com>
Cc: Tom Lane <tgl@sss.pgh.pa.us>, Andreas Seltenreich <seltenreich@gmx.de>, Pg Hackers <pgsql-hackers@postgresql.org>, Simon Riggs <simon@2ndquadrant.com>
Date: 2017-04-10T05:14:42Z
Lists: pgsql-hackers 
>>>>> "Thomas" == Thomas Munro <thomas.munro@enterprisedb.com> writes:

 >> SomeType *x = (SomeType *) lfirst(l);
 >> 
 >> (in my code I tend to omit the (SomeType *), which I dislike because
 >> it adds no real protection)

 Thomas> Just BTW, without that cast it's not compilable as C++, so I'm
 Thomas> guessing that Peter E will finish up putting it back in
 Thomas> wherever you leave it out...

There's north of 150 other examples (just grep for '= lfirst' in the
source). Some were even committed by Peter E :-)

In the discussion with Andres the same point came up for palloc, for
which I suggested we add something along the lines of:

#define palloc_object(_type_) (_type_ *) palloc(sizeof(_type_))
#define palloc_array(_type_, n) (_type_ *) palloc((n) * sizeof(_type_))

palloc() without a cast is even more common than lfirst() without one,
and something like half of those (and 80%+ of the pallocs that do have a
cast) are palloc(sizeof(...)) or palloc(something * sizeof(...)).

-- 
Andrew (irc:RhodiumToad)

Commits

  1. Improve castNode notation by introducing list-extraction-specific variants.

  2. Clean up bugs in clause_selectivity() cleanup.

  3. Add newly-symlinked files to "make clean" target.

  4. Reset API of clause_selectivity()