Re: Win32 testing needed

Andreas Pflug <pgadmin@pse-consulting.de> writes:
> Maybe we should consider examining GetLastError() and FormatMessage()
> (the equivalent of strerror, a sample is in the symlink/junction code)
> for %m under win32; these will work for standard posix operations also
> and might give much more detailed messages in many situations.

Hmm.  It's a bit attractive but it would break a lot of existing code
that assumes saving/restoring errno is a sufficient way to not clobber
the error result info between getting an error and reporting it.  For
instance, xlog.c has several repetitions of this pattern:

        errno = 0;
        if ((int) write(fd, zbuffer, sizeof(zbuffer)) != (int) sizeof(zbuffer))
        {
            int            save_errno = errno;

            /*
             * If we fail to make the file, delete it to release disk
             * space
             */
            unlink(tmppath);
            /* if write didn't set errno, assume problem is no disk space */
            errno = save_errno ? save_errno : ENOSPC;

            ereport(PANIC,
                    (errcode_for_file_access(),
                     errmsg("could not write to file \"%s\": %m", tmppath)));
        }

and I don't think I want to add Windows #ifdefs to every such place.

What would be more supportable is something that assigns a suitable
value to errno after a failure of a Windows-specific call.  Thus
something like

        if (!ReadFile(syslogPipe[0], logbuffer, sizeof(logbuffer),
                      &bytesRead, 0))
        {
            DWORD error = GetLastError();

            if (error == ERROR_HANDLE_EOF ||
                error == ERROR_BROKEN_PIPE)
                break;
            errno = Win32ErrorToErrno(error);
            ereport(LOG,
                    (errcode_for_file_access(),
                     errmsg("could not read from logger pipe: %m")));
        }

While this is admittedly ugly, it confines the ugliness to the fairly
small number of places where we call Windows-specific routines (ie,
we're already inside an #ifdef WIN32).  If we do the other, the
messiness is going to spread into what had been perfectly good
Posix-standard code.

            regards, tom lane