Thread: pgsql: Recognize ERROR_SHARING_VIOLATION (translate to EACCES), increase

Log Message:
-----------
Recognize ERROR_SHARING_VIOLATION (translate to EACCES), increase log
level for unrecognized win32 error codes to LOG, and make messages
conform to style guide.  Per old suggestion from Qingqing Zhou, which
seems to have gotten lost in the shuffle.

Modified Files:
--------------
    pgsql/src/backend/port/win32:
        error.c (r1.4 -> r1.5)
        (http://developer.postgresql.org/cvsweb.cgi/pgsql/src/backend/port/win32/error.c.diff?r1=1.4&r2=1.5)

"Tom Lane" <tgl@svr1.postgresql.org> wrote
> Recognize ERROR_SHARING_VIOLATION (translate to EACCES), increase log
> level for unrecognized win32 error codes to LOG, and make messages
> conform to style guide.

This patch does not really work. This is because when mdread()/mdwrite()
fails, it does *not* call our own _dosmaperr() routine. Instead, it calls
the Win32 CRT's _dosmapperr() to do the translation (which fails to
translate ERROR_SHARING_VIOLATION either). If we want our _dosmaperr() to
play role in, we have to call it explicitely in mdread()/mdwrite() or
lower levels like FileRead()/FileWrite().

I would suggest don't revert this patch. When we get the feedback from
Jean-Pierre Pelletier, we can make the patch complete.

Regards,
Qingqing

"Tom Lane" <tgl@svr1.postgresql.org> wrote
> Recognize ERROR_SHARING_VIOLATION (translate to EACCES), increase log
> level for unrecognized win32 error codes to LOG, and make messages
> conform to style guide.

This patch does not really work. This is because when mdread()/mdwrite()
fails, it does *not* call our own _dosmaperr() routine. Instead, it calls
the Win32 CRT's _dosmapperr() to do the translation (which fails to
translate ERROR_SHARING_VIOLATION either). If we want our _dosmaperr() to
play role in, we have to call it explicitely in mdread()/mdwrite() or lower
levels like FileRead()/FileWrite().

I would suggest don't revert this patch. When we get the feedback from
Jean-Pierre Pelletier, we can make the patch complete.

Regards,
Qingqing

Qingqing Zhou <zhouqq@cs.toronto.edu> writes:
> "Tom Lane" <tgl@svr1.postgresql.org> wrote
>> Recognize ERROR_SHARING_VIOLATION (translate to EACCES), increase log
>> level for unrecognized win32 error codes to LOG, and make messages
>> conform to style guide.

> This patch does not really work. This is because when mdread()/mdwrite()
> fails, it does *not* call our own _dosmaperr() routine.

Agreed, it does not help mdread, but it's still an improvement.

            regards, tom lane