Horton

Re: Thread-unsafe coding in ecpg

Tom Lane <tgl@sss.pgh.pa.us>

View thread View on postgr.es
From: Tom Lane <tgl@sss.pgh.pa.us>
To: Andrew Gierth <andrew@tao11.riddles.org.uk>
Cc: Michael Meskes <meskes@postgresql.org>, pgsql-hackers@lists.postgresql.org
Date: 2019-01-20T02:00:37Z
Lists: pgsql-hackers 
Andrew Gierth <andrew@tao11.riddles.org.uk> writes:
> Would it help if we had non-locale-aware functions for both
> floating-point output _and_ input? i.e. import a known-working strtod()
> (allowing us to remove all the hacks that have grown up around it, for
> special-case input and wonky error handling) with locale functionality
> removed.

Dunno, is there such a thing as a platform-independent strtod()?
I'd have thought that, for instance, typical implementations would
be pretty much in bed with the details of IEEE float format ---
your example where strtof() is different from (float) strtod()
makes it hard to believe that it can be written without assumptions
about the hardware's float format.

(Note that this concern is independent of whether we adopt the Ryu
code, which IIUC also depends on IEEE floats.  Our answer for anyone
wanting to run on non-IEEE hardware can be to #ifdef out Ryu and use
the existing float output code.  But doing the equivalent thing on the
input side wouldn't solve ecpg's problem.)

			regards, tom lane

Commits

  1. Remove _configthreadlocale() calls in ecpg test suite.

  2. Remove infinite-loop hazards in ecpg test suite.

  3. Blind attempt to fix _configthreadlocale() failures on MinGW.

  4. Avoid thread-safety problem in ecpglib.

  5. Second try at fixing ecpglib thread-safety problem.