Обсуждение: pg_ctl
I briefly reviewed the alpha version of this that Joshua posted a URL to (after converting all the c++ style comments and running through pgindent so I could read it). I endorse Neil's and Bruce's comments. In addition, I noticed these things: . on Win32 a start op is guaranteed to hang unless I'm misreading the code, as it is not backgrounded. Calling system() won't work here - I think we'll have to call CreateProcess() on Win32. . what about devnull processing on Win32? . multi-line literals are apparently illegal (see recent hackers thread?) and rejected by some compilers - use the fact that the compiler will concatentate adjacent string literals. . there is a lot of incorrect code that tries to do things that are already done correctly and portably in initdb.c - e.g. finding the postmaster executable. . calling fopen() to check existence of a file is, er, suboptimal. Use stat(). . no error checking on malloc() calls . calling free() on these tiny strings is quite unnecessary - the program should use minimal memory and have a very short life span. It's just useless clutter IMNSHO. This isn't an exhaustive list. I suspect this work is actually a bit premature, and would best be done after we have a working Win32 postmaster. After all, you can happily run postgres without ever calling pg_ctl - it's a convenience rather than an absolute necessity (unlike initdb). On Windows, production users probably wouldn't use it at all - they would use the service manager, I should think. cheers andrew
Communicated. Andrew Dunstan wrote: > > I briefly reviewed the alpha version of this that Joshua posted a URL > to (after converting all the c++ style comments and running through > pgindent so I could read it). > > I endorse Neil's and Bruce's comments. In addition, I noticed these > things: > > . on Win32 a start op is guaranteed to hang unless I'm misreading the > code, as it is not backgrounded. Calling system() won't work here - I > think we'll have to call CreateProcess() on Win32. > . what about devnull processing on Win32? > . multi-line literals are apparently illegal (see recent hackers > thread?) and rejected by some compilers - use the fact that the > compiler will concatentate adjacent string literals. > . there is a lot of incorrect code that tries to do things that are > already done correctly and portably in initdb.c - e.g. finding the > postmaster executable. > . calling fopen() to check existence of a file is, er, suboptimal. Use > stat(). > . no error checking on malloc() calls > . calling free() on these tiny strings is quite unnecessary - the > program should use minimal memory and have a very short life span. > It's just useless clutter IMNSHO. > > This isn't an exhaustive list. > > I suspect this work is actually a bit premature, and would best be > done after we have a working Win32 postmaster. After all, you can > happily run postgres without ever calling pg_ctl - it's a convenience > rather than an absolute necessity (unlike initdb). On Windows, > production users probably wouldn't use it at all - they would use the > service manager, I should think. > > cheers > > andrew > > > > > > ---------------------------(end of broadcast)--------------------------- > TIP 1: subscribe and unsubscribe commands go to majordomo@postgresql.org -- Command Prompt, Inc., home of Mammoth PostgreSQL - S/ODBC and S/JDBC Postgresql support, programming shared hosting and dedicated hosting. +1-503-667-4564 - jd@commandprompt.com - http://www.commandprompt.com Mammoth PostgreSQL Replicator. Integrated Replication for PostgreSQL
Back on December 1st I wrote: > > > I suspect this work is actually a bit premature, and would best be > done after we have a working Win32 postmaster. After all, you can > happily run postgres without ever calling pg_ctl - it's a convenience > rather than an absolute necessity (unlike initdb). On Windows, > production users probably wouldn't use it at all - they would use the > service manager, I should think. > > Are we now in a state where it would make sense to look at doing pg_ctl in C? (And has anyone looked at putting in service manager hooks?) cheers andrew
Andrew Dunstan <andrew@dunslane.net> writes:
> Are we now in a state where it would make sense to look at doing pg_ctl
> in C?
Why bother? I should think that Windows users want to run PG as a service.
regards, tom lane
Tom Lane wrote: >Andrew Dunstan <andrew@dunslane.net> writes: > > >>Are we now in a state where it would make sense to look at doing pg_ctl >>in C? >> >> > >Why bother? I should think that Windows users want to run PG as a service. > > Not unreasonable. Maybe someone who uses/distributes a commercial Windows version can inform us about usage patterns. If that's what we decide we should remove this from the Win32 page: . Create C version of pg_ctl? And while on the subject of that page, this issue should now be fixed with the listen_addresses patch (which I hope someone can verify): Postmaster must be started with -i flag, or with tcpip_socket = true in postgresql.conf, else the postmaster will fail to start and issue the following error: |LOG: select() failed in postmaster: No such file or directory cheers andrew |
> Not unreasonable. Maybe someone who uses/distributes a commercial > Windows version can inform us about usage patterns. If that's what we > decide we should remove this from the Win32 page: > > . Create C version of pg_ctl? > Well we are (although it is on the back burner) working on pg_ctl in C. Our experience is that all of our customers who run our Cygwin based version prefer it running as a service. > And while on the subject of that page, this issue should now be fixed > with the listen_addresses patch (which I hope someone can verify): > > Postmaster must be started with -i flag, or with tcpip_socket = true > in postgresql.conf, else the postmaster will fail to start and issue > the following error: > |LOG: select() failed in postmaster: No such file or directory > > > cheers > > > andrew > | > > ---------------------------(end of broadcast)--------------------------- > TIP 9: the planner will ignore your desire to choose an index scan if > your > joining column's datatypes do not match -- Command Prompt, Inc., home of Mammoth PostgreSQL - S/ODBC and S/JDBC Postgresql support, programming shared hosting and dedicated hosting. +1-503-667-4564 - jd@commandprompt.com - http://www.commandprompt.com PostgreSQL Replicator -- production quality replication for PostgreSQL
Joshua D. Drake wrote:
>
> > Not unreasonable. Maybe someone who uses/distributes a commercial
> > Windows version can inform us about usage patterns. If that's what we
> > decide we should remove this from the Win32 page:
> >
> > . Create C version of pg_ctl?
> >
> Well we are (although it is on the back burner) working on pg_ctl in C.
>
> Our experience is that all of our customers who run our Cygwin based
> version prefer it running as a service.
We should make pg_ctl a C program. Some package installer or something
will need it. It is only 412 lines of shell script.
> > And while on the subject of that page, this issue should now be fixed
> > with the listen_addresses patch (which I hope someone can verify):
> >
> > Postmaster must be started with -i flag, or with tcpip_socket = true
> > in postgresql.conf, else the postmaster will fail to start and issue
> > the following error:
> > |LOG: select() failed in postmaster: No such file or directory
Win32 project page updated:
http://momjian.postgresql.org/main/writings/pgsql/project
--
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
Tom Lane wrote: >>Are we now in a state where it would make sense to look at doing pg_ctl >>in C? > > Why bother? I should think that Windows users want to run PG as a service. Except when a non administrator wants to run his own PG instance. -- dave
> >>Are we now in a state where it would make sense to look at doing > >>pg_ctl > >>in C? > > > > Why bother? I should think that Windows users want to run PG as a > > service. > > Except when a non administrator wants to run his own PG instance. This is one of the reasons why a pg_ctl is still needed. The other is if you run the server in debugging on the commandline. You can shut it down with Ctrl-C, but you can't send signals to reload etc. //Magnus
> Are we now in a state where it would make sense to look at > doing pg_ctl in C? FWIW, I think the Service Manager integration is crucial, and pg_ctl in C is a nice to have. > (And has anyone looked at putting in service manager hooks?) I did for a private port, but haven't fixed it up yet for this port. If anyone wants to take this item, just let me know (I can send you the code I've got already). Cheers, Claudio --- Certain disclaimers and policies apply to all email sent from Memetrics. For the full text of these disclaimers and policies see <a href="http://www.memetrics.com/emailpolicy.html">http://www.memetrics.com/em ailpolicy.html</a>