"Ross J. Reedstrom" <reedstrm(at)rice(dot)edu> writes:
> Reading into this further, it seems to me that this code is attempting to
> implement DLL dependencies. I'd think that the DLL loader is _supposed_
> to handle that automatically (based on te docmuentation for LoadLibrary
> mentioning that it can cause dependent libraries to also be loaded), but
> I know even less about building win32 dlls than I do about coding them.
Ditto. But, looking at the cvs logs, it seems that the use of netmsg
was inserted rather hastily, and it may well be done completely the
I'm not sure that declaring a dependency to the DLL loader is the way we
want to go though; wouldn't that cause the system to refuse to run libpq
at all unless netmsg was available? That's certainly not the behavior
we want, considering that netmsg is only a fallback source of error
strings in the first place, and that we can fall back to "no error text
available" if we can't load it.
Loading and unloading on-the-fly in winsock_strerror is looking better
and better to me.
regards, tom lane
In response to
pgsql-interfaces by date
|Next:||From: Hiroshi Inoue||Date: 2001-11-27 23:36:30|
|Subject: Re: ODBC driver with dynamic cursor support ??|
|Previous:||From: Tom Lane||Date: 2001-11-27 19:29:13|
|Subject: Re: Can a windows DLL have more than one process attached? |