Re: ecpg threading vs win32

On Mon, Mar 19, 2007 at 09:33:54AM +0900, ITAGAKI Takahiro wrote:
>
> Magnus Hagander <magnus(at)hagander(dot)net> wrote:
>
> > This patch replaces the pthreads code in ecpg with native win32 threads,
> > in order to make it threadsafe. The idea is not to have to download the
> > non-standard pthreads library on windows.
> >
> > Does it seem like it should be doing the right thing? Does somebody have
> > a good test-case where ecpg breaks when not built thread-safe? (which
> > would then also break when built thread-safe with a broken implementation)
>
> I have two questions about thread-safe ecpg.
>
> Q1. Don't you use CRITICAL_SECTION instead of Mutex (CreateMutex)?
> I've heard there is a performance benefit in CRITICAL_SECTION.
> If the mutex is shared only in one process, CS might be a better solution.
> http://japan.internet.com/developer/img/article/873/17801.gif
> http://world.std.com/~jmhart/csmutx.htm

Yes, CS can be slightly faster. Though under this use-pattern, I think
it will not make a measurable difference at all.
The reason I went with Mutex is that I wanted it to be as similar as
possible to the pthreads code.

> Q2. Do we need to use PQescapeStringConn() instead of PQescapeString()?
> PQescapeString() is used to escape literals, and the documentation says
> PQescapeStringConn() should be used in multi-threaded client programs.
> http://momjian.us/main/writings/pgsql/sgml/libpq-exec.html#LIBPQ-EXEC-ESCAPE-STRING
> | PQescapeString can be used safely in single-threaded client programs
> | that work with only one PostgreSQL connection at a time

Seems so, but that's unrelated to this patch ;-) I'll leave the final
comment on that up to Michael.

//Magnus