Обсуждение: Re: REPOST: InitDB Failure on install

Поиск
Список
Период
Сортировка

Re: REPOST: InitDB Failure on install

От
"Magnus Hagander"
Дата:
>>>> (BTW, there was some discussion around this point while
>you were gone.
>>>> Dunno if you've caught up on the list traffic yet.)
>
>> Ok. I think I'm caught up now, and I don't find anything directly
>> related to this. Could be because there was so much mail
>that I skimmed
>> some of it a bit too quickly... Could you provide me with a pointer?
>
>It took some digging, but here is the thread I had in mind:
>http://archives.postgresql.org/pgsql-hackers-win32/2004-08/msg00047.php
>http://archives.postgresql.org/pgsql-hackers-win32/2004-08/msg00050.php
>
>Life does get simpler if we can just do "errno = GetLastError()" but
>I strongly suspect we're gonna end up wanting to do some mapping.

Arrgh. I did see that one, just didn't connect. I guess the brain gets
mushy after about 2500 mails.. Sorry about that one.


Looking around a bit on the net, I see tons and tons of examples of
people just doing the "errno=GetLastError()" codepath. And I find no
evidence anywhere that this should not be safe. For winsock, look at for
example
http://msdn.microsoft.com/library/default.asp?url=/library/en-us/winsock
/winsock/error_codes_errno_h_errno_and_wsagetlasterror_2.asp which
states "For compatibility with BSD, an application may choose to include
a line of the form: #define errno WSAGetLastError". And
WSAGetLastError() is equivalent to GetLastError() in any windows these
days.

Based on this, I think we're quite safe doing errno=GetLastError() in
any places where we need it.

What we might need to do is change the useful_strerror() function to use
the FormatMessage function for all messages on win32, and not as now is
done just for winsock functions. I'm not 100% sure this is needed - it's
possible it handles all the standard messages and just not winsock. But
it should not report *incorrect* error messages, it would just not know
the text for the error message.


In order to be future safe we ight want to do something like put a macro
in win32.h doing
#define Win32ErrorToErrno(i) i
and then use
errno=Win32ErrorToErrno(GetLastError())

In case we ever need a mapping, we can just change that macro into a
function call, without needing to revisit all the code. Not sure if it's
worth it, though?

//Magnus


Re: REPOST: InitDB Failure on install

От
Tom Lane
Дата:
"Magnus Hagander" <mha@sollentuna.net> writes:
> In order to be future safe we ight want to do something like put a macro
> in win32.h doing
> #define Win32ErrorToErrno(i) i

That was what I was gonna suggest, just in case.

            regards, tom lane

Re: REPOST: InitDB Failure on install

От
Bruce Momjian
Дата:
Added to open items:

            o improve ereport %m error messages


---------------------------------------------------------------------------

Magnus Hagander wrote:
> >>>> (BTW, there was some discussion around this point while
> >you were gone.
> >>>> Dunno if you've caught up on the list traffic yet.)
> >
> >> Ok. I think I'm caught up now, and I don't find anything directly
> >> related to this. Could be because there was so much mail
> >that I skimmed
> >> some of it a bit too quickly... Could you provide me with a pointer?
> >
> >It took some digging, but here is the thread I had in mind:
> >http://archives.postgresql.org/pgsql-hackers-win32/2004-08/msg00047.php
> >http://archives.postgresql.org/pgsql-hackers-win32/2004-08/msg00050.php
> >
> >Life does get simpler if we can just do "errno = GetLastError()" but
> >I strongly suspect we're gonna end up wanting to do some mapping.
>
> Arrgh. I did see that one, just didn't connect. I guess the brain gets
> mushy after about 2500 mails.. Sorry about that one.
>
>
> Looking around a bit on the net, I see tons and tons of examples of
> people just doing the "errno=GetLastError()" codepath. And I find no
> evidence anywhere that this should not be safe. For winsock, look at for
> example
> http://msdn.microsoft.com/library/default.asp?url=/library/en-us/winsock
> /winsock/error_codes_errno_h_errno_and_wsagetlasterror_2.asp which
> states "For compatibility with BSD, an application may choose to include
> a line of the form: #define errno WSAGetLastError". And
> WSAGetLastError() is equivalent to GetLastError() in any windows these
> days.
>
> Based on this, I think we're quite safe doing errno=GetLastError() in
> any places where we need it.
>
> What we might need to do is change the useful_strerror() function to use
> the FormatMessage function for all messages on win32, and not as now is
> done just for winsock functions. I'm not 100% sure this is needed - it's
> possible it handles all the standard messages and just not winsock. But
> it should not report *incorrect* error messages, it would just not know
> the text for the error message.
>
>
> In order to be future safe we ight want to do something like put a macro
> in win32.h doing
> #define Win32ErrorToErrno(i) i
> and then use
> errno=Win32ErrorToErrno(GetLastError())
>
> In case we ever need a mapping, we can just change that macro into a
> function call, without needing to revisit all the code. Not sure if it's
> worth it, though?
>
> //Magnus
>
>
> ---------------------------(end of broadcast)---------------------------
> TIP 8: explain analyze is your friend
>

--
  Bruce Momjian                        |  http://candle.pha.pa.us
  pgman@candle.pha.pa.us               |  (610) 359-1001
  +  If your life is a hard drive,     |  13 Roberts Road
  +  Christ can be your backup.        |  Newtown Square, Pennsylvania 19073