Postgres Pro
Downloads
Home   >   mailing lists

Thread: bsd indent

bsd indent

From
Andrew Dunstan
Date:
I grabbed the patched BSD indent from a mirror. Here's a couple of points:

. with a little trial and error GNU indent 2.2.9 actually did quite a
reasonable job for me last night. But I guess YMMV. I'm curious to know
what nasty mangling it does.
. the file is a compressed tar file - it would be better to have a file
extension of .tgz or .tar.gz  instead of just .gz, so that this is apparent.
. I had to hack the makefile for my linux box for a reasonably clean
build, so that the CFLAGS on the install line doesn't override setting
the defines, and the compiler won't complain about isblank and/or strdup
and/or fileno. The diff is attached for anyone interested.


cheers

andrew
4d3
< #
6,7c5,6
< XFLAGS =
< CFLAGS = -g -Wall -D__RCSID="static char *rcsid=" -D__COPYRIGHT="static char *copyright="
---
> XFLAGS = -D_GNU_SOURCE -D__RCSID="static char *rcsid=" -D__COPYRIGHT="static char *copyright="
> CFLAGS = -g -Wall
11c10
<     $(CC) -o $(TARGET) $(XFLAGS) $(CFLAGS) args.o indent.o io.o lexi.o parse.o pr_comment.o $(LIBS)
---
>     $(CC) -o $(TARGET) $(CFLAGS) args.o indent.o io.o lexi.o parse.o pr_comment.o $(LIBS)

Re: bsd indent

From
Bruce Momjian
Date:
Andrew Dunstan wrote:
> 
> I grabbed the patched BSD indent from a mirror. Here's a couple of points:
> 
> . with a little trial and error GNU indent 2.2.9 actually did quite a 
> reasonable job for me last night. But I guess YMMV. I'm curious to know 
> what nasty mangling it does.

GNU indent does a reasonable job, but it did some strange spacing for
function definitions, if I remember correctly.  Can you test by running
pgindent and gnuindent through the /executor files and do a diff to see
the differences in formatting?

> . the file is a compressed tar file - it would be better to have a file 
> extension of .tgz or .tar.gz  instead of just .gz, so that this is apparent.

Done.  Sorry.

> . I had to hack the makefile for my linux box for a reasonably clean 
> build, so that the CFLAGS on the install line doesn't override setting 
> the defines, and the compiler won't complain about isblank and/or strdup 
> and/or fileno. The diff is attached for anyone interested.

OK, I updated it on the ftp site.  Can you download it?  It should have
size 41192.  Is _GNU_SOURCE required?  I didn't add that, but fixed the
install part.

--  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,
Pennsylvania19073

Re: bsd indent

From
Andrew Dunstan
Date:
Bruce Momjian wrote:

>Andrew Dunstan wrote:
>  
>
>>I grabbed the patched BSD indent from a mirror. Here's a couple of points:
>>
>>. with a little trial and error GNU indent 2.2.9 actually did quite a 
>>reasonable job for me last night. But I guess YMMV. I'm curious to know 
>>what nasty mangling it does.
>>    
>>
>
>GNU indent does a reasonable job, but it did some strange spacing for
>function definitions, if I remember correctly.  Can you test by running
>pgindent and gnuindent through the /executor files and do a diff to see
>the differences in formatting?
>

after I added -ncdb to the GNU indent options in pgindent, the main 
effect appears to be that it joins function calls across lines, 
sometimes in weird ways. I guess that's bad enough for us not to use it 
unless someone knows a way to turn that off.

Example (second case is bsd indented):

***************
*** 217,223 **** execTuplesHashPrepare(TupleDesc tupdesc,                     int numCols,
AttrNumber*matchColIdx,
 
!                     FmgrInfo **eqfunctions, FmgrInfo **hashfunctions) {   int         i;
--- 234,241 ---- execTuplesHashPrepare(TupleDesc tupdesc,                     int numCols,
AttrNumber*matchColIdx,
 
!                     FmgrInfo **eqfunctions,
!                     FmgrInfo **hashfunctions) {   int         i;


>>. I had to hack the makefile for my linux box for a reasonably clean 
>>build, so that the CFLAGS on the install line doesn't override setting 
>>the defines, and the compiler won't complain about isblank and/or strdup 
>>and/or fileno. The diff is attached for anyone interested.
>>    
>>
>
>OK, I updated it on the ftp site.  Can you download it?  It should have
>size 41192.  Is _GNU_SOURCE required?  I didn't add that, but fixed the
>install part.
>
>  
>

I am being refused connection to developer.postgresql.org and it hasn't 
reached the mirrors yet AFAICS.

Here's what happens on my RH9 box if you don't define _GNU_SOURCE:
[andrew@Thor indent]$ make clean
rm -f *.o indent log core
[andrew@Thor indent]$ make XFLAGS='-D__RCSID="static char *rcsid=" 
-D__COPYRIGHT="static char *copyright="'
cc -c -D__RCSID="static char *rcsid=" -D__COPYRIGHT="static char 
*copyright=" -g -Wall args.c
args.c:71: warning: `rcsid' defined but not used
cc -c -D__RCSID="static char *rcsid=" -D__COPYRIGHT="static char 
*copyright=" -g -Wall indent.c
indent.c:68: warning: `copyright' defined but not used
indent.c:78: warning: `rcsid' defined but not used
cc -c -D__RCSID="static char *rcsid=" -D__COPYRIGHT="static char 
*copyright=" -g -Wall io.c
io.c:71: warning: `rcsid' defined but not used
cc -c -D__RCSID="static char *rcsid=" -D__COPYRIGHT="static char 
*copyright=" -g -Wall lexi.c
lexi.c:71: warning: `rcsid' defined but not used
cc -c -D__RCSID="static char *rcsid=" -D__COPYRIGHT="static char 
*copyright=" -g -Wall parse.c
parse.c:71: warning: `rcsid' defined but not used
cc -c -D__RCSID="static char *rcsid=" -D__COPYRIGHT="static char 
*copyright=" -g -Wall pr_comment.c
pr_comment.c: In function `pr_comment':
pr_comment.c:409: warning: implicit declaration of function `isblank'
pr_comment.c: At top level:
pr_comment.c:71: warning: `rcsid' defined but not used
cc -o indent -g -Wall args.o indent.o io.o lexi.o parse.o pr_comment.o
[andrew@Thor indent]$ make clean
rm -f *.o indent log core
[andrew@Thor indent]$ make
cc -c -D_GNU_SOURCE -D__RCSID="static char *rcsid=" 
-D__COPYRIGHT="static char *copyright=" -g -Wall args.c
args.c:71: warning: `rcsid' defined but not used
cc -c -D_GNU_SOURCE -D__RCSID="static char *rcsid=" 
-D__COPYRIGHT="static char *copyright=" -g -Wall indent.c
indent.c:68: warning: `copyright' defined but not used
indent.c:78: warning: `rcsid' defined but not used
cc -c -D_GNU_SOURCE -D__RCSID="static char *rcsid=" 
-D__COPYRIGHT="static char *copyright=" -g -Wall io.c
io.c:71: warning: `rcsid' defined but not used
cc -c -D_GNU_SOURCE -D__RCSID="static char *rcsid=" 
-D__COPYRIGHT="static char *copyright=" -g -Wall lexi.c
lexi.c:71: warning: `rcsid' defined but not used
cc -c -D_GNU_SOURCE -D__RCSID="static char *rcsid=" 
-D__COPYRIGHT="static char *copyright=" -g -Wall parse.c
parse.c:71: warning: `rcsid' defined but not used
cc -c -D_GNU_SOURCE -D__RCSID="static char *rcsid=" 
-D__COPYRIGHT="static char *copyright=" -g -Wall pr_comment.c
pr_comment.c:71: warning: `rcsid' defined but not used
cc -o indent -g -Wall args.o indent.o io.o lexi.o parse.o pr_comment.o
[andrew@Thor indent]$


cheers

andrew

Re: bsd indent

From
Bruce Momjian
Date:
Andrew Dunstan wrote:
> >GNU indent does a reasonable job, but it did some strange spacing for
> >function definitions, if I remember correctly.  Can you test by running
> >pgindent and gnuindent through the /executor files and do a diff to see
> >the differences in formatting?
> >
> 
> after I added -ncdb to the GNU indent options in pgindent, the main 
> effect appears to be that it joins function calls across lines, 
> sometimes in weird ways. I guess that's bad enough for us not to use it 
> unless someone knows a way to turn that off.
> 
> Example (second case is bsd indented):
> 
> ***************
> *** 217,223 ****
>   execTuplesHashPrepare(TupleDesc tupdesc,
>                       int numCols,
>                       AttrNumber *matchColIdx,
> !                     FmgrInfo **eqfunctions, FmgrInfo **hashfunctions)
>   {
>     int         i;
>  
> --- 234,241 ----
>   execTuplesHashPrepare(TupleDesc tupdesc,
>                       int numCols,
>                       AttrNumber *matchColIdx,
> !                     FmgrInfo **eqfunctions,
> !                     FmgrInfo **hashfunctions)
>   {
>     int         i;
>  

Yes, this is exactly what I remember, and I can think of no way to
"script" that fix.

Let me know when you can download.  I sure it will hit the mirrors
shortly.

The fact that we have a way to auto-indent 450k lines of C code with
few/no complaints is amazing in itself.

--  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,
Pennsylvania19073

Re: bsd indent

From
Andrew Dunstan
Date:
BTW, there's a bug in entab.c, too. This line:

#include "../include/c.h"

should read

#include "../../include/c.h"

cheers

andrew

Re: bsd indent

From
Bruce Momjian
Date:
Andrew Dunstan wrote:
> 
> BTW, there's a bug in entab.c, too. This line:
> 
> #include "../include/c.h"
> 
> should read
> 
> #include "../../include/c.h"

Thanks.  Fixed.

--  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,
Pennsylvania19073