Re: [PATCHES] pg_ctl

From: Neil Conway <neilc(at)samurai(dot)com>
To: Bruce Momjian <pgman(at)candle(dot)pha(dot)pa(dot)us>
Cc: "Joshua D(dot) Drake" <jd(at)commandprompt(dot)com>, PostgreSQL Win32 port list <pgsql-hackers-win32(at)postgresql(dot)org>, PostgreSQL-patches <pgsql-patches(at)postgresql(dot)org>
Subject: Re: [PATCHES] pg_ctl
Date: 2003-11-29 00:10:47
Message-ID: 87smk8japk.fsf@mailbox.samurai.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers-win32 pgsql-patches

This code is pretty awful, IMHO.

If you're going to copy code from a 3rd party (in this case, MSDN), it
is standard practise to include an attribution. Also, what
redistribution terms apply to MSDN sample code?

Assuming we don't remove or rewrite the MSDN sample code, it is
usually considered good practise to put imported code into a separate
file.

You can remove the kludge for 16-bit processes, as PostgreSQL won't
ever be one. Dunno if it's worth modifying the sample code for this.

It is generally considered good practise to divide distinct
functionality into separate functions, rather than use a ~450 line
main() function.

I'd like to see fewer WIN32 #ifdefs in the main code path; please also
endeavour to make them as small as possible.

Copying and pasting code from one branch of an #ifdef or if into the
other should also be avoided where possible: this is done in a couple
places.

Other random badness I noticed while browsing through the code:

#ifdef WIN32
char *bindir="c:\\pgsql\\bin";
char *VERSION="7.3.5";
#else
char *bindir="@bindir@";
char *VERSION="@VERSION@";
#endif

This is plainly wrong.

char* DEFPOSTOPTS= NULL;
char* POSTOPTSFILE= NULL;
char* PIDFILE= NULL;

int n_sig;

char *logfile=NULL;
char *silence_echo=NULL;
char *shutdown_mode="smart";
char *op=NULL;
int PID;

Can we please try to follow *some* kind of coherent convention for
naming and indentation?

char* buffer;
char* buffer2;

Please put some thought into your variable names.

po_path=(char*)malloc(sizeof(char)* (strlen(PGPATH)+strlen(pgsqlexe)+2));
sprintf(po_path, "%s%s%s",PGPATH, path_delim, pgsqlexe);

This code needlessly assumes strlen(path_delim) == 1. On the other
hand, it is reasonable to assume sizeof(char) == 1 (this is guaranteed
by ANSI C).

for ( ;cont; token = (char*)strtok(NULL, env_delim)){
if (token==NULL) cont=0; else {
buffer=(char*)malloc(sizeof(char)* (strlen(token)+strlen(CMDNAME)+1));
sprintf(buffer, token);
strcat (buffer, path_delim);
strcat (buffer, CMDNAME);
if (post_file=fopen(buffer, "r")){
self_path=token;
cont=0;
free(buffer);
} else free(buffer);
}
if (token==NULL) cont=0;
}

Good lord.

-Neil

In response to

  • Re: pg_ctl at 2003-11-28 22:20:37 from Bruce Momjian

Responses

Browse pgsql-hackers-win32 by date

  From Date Subject
Next Message Neil Conway 2003-11-29 00:25:58 Re: [PATCHES] pg_ctl
Previous Message Bruce Momjian 2003-11-28 22:20:37 Re: pg_ctl

Browse pgsql-patches by date

  From Date Subject
Next Message Neil Conway 2003-11-29 00:25:58 Re: [PATCHES] pg_ctl
Previous Message Bruce Momjian 2003-11-28 22:20:37 Re: pg_ctl