sed and configure scripts

sed and configure scripts

[Ville Herva]

I must be doing something wrong here. Basicly, it seems like sed would
always intepret its input as binary.

I'm trying to run zsh-3.1.6 configure script. All is fine until configure
does something like

cat > c.sed << EOF
s/ristiina/parkano/g
s/oulu/turku/g
EOF
sed -n -f c.sed < naantali > hollola

Now, c.sed contains dos newlines (0x0d), and sed barfs:

sed: file conftest.frag line 1: Unknown option to 's'

(option 'g^M'...)

my mounts are textmode, but the situation did not change with binary
either. The configure script (the cat > ... and the following lines in
particular) has no ^M's. CYGWIN=binmode is _not_ set.

cygwin1.ddl snapshot 15052000, newest sed from net.

I understand somebody (Earnie?) already compiled zsh? Is there anything
else to watch out? s/__CYGWIN/__CYGWIN__/, but where?

Hmm. I wish the Amol Deshpande's native zsh would work 'right' with
cygwin. Not only, because I can't seem to compile zsh, but the native
zsh has some neat features...


-- v --

v@iki.fi

--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

Re: sed and configure scripts

[Ville Herva]

On Mon, May 29, 2000 at 11:59:11AM +0300, you [Ville Herva] claimed:
> I must be doing something wrong here. Basicly, it seems like sed would
> always intepret its input as binary.
> 
> I'm trying to run zsh-3.1.6 configure script. All is fine until configure
> does something like
> 
> cat > c.sed << EOF
> s/ristiina/parkano/g
> s/oulu/turku/g
> EOF
> sed -n -f c.sed < naantali > hollola
> 
> Now, c.sed contains dos newlines (0x0d), and sed barfs:
> 
> sed: file conftest.frag line 1: Unknown option to 's'
> 
> (option 'g^M'...)
> 
> my mounts are textmode, but the situation did not change with binary
> either. The configure script (the cat > ... and the following lines in
> particular) has no ^M's. CYGWIN=binmode is _not_ set.
> 
> cygwin1.ddl snapshot 15052000, newest sed from net.

This is a problem with sed. I tried to recompile sed but its configure
script failed the same way! How do you people get configure scripts run? I
couldn't even figure out any bearable work-around.

Sed uses mmap to read executed file:

sed.c::main():
        case 'f':
          the_program = compile_file(the_program, optarg);
          break;

compile.c:.compile_file():
  if (cmdfile[0] != '-' || cmdfile[1] != '\0')
    prog.file = ck_fopen(cmdfile, "r");
  if (map_file(prog.file, &prog.base, &len))
  
  
util.c::cd_fopen():
  if ( ! (fp = fopen(name, mode)) )
  
  
sed.c::map_file():
      nbase = VCAST(VOID *)mmap(NULL, CAST(size_t)s.st_size, PROT_READ,
                                MAP_PRIVATE, fileno(fp), CAST(off_t)0);
                                

I assume CR/LF translation is not done for mmap? 

Am I missing something obvious?


-- v --

v@iki.fi

--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

Re: sed and configure scripts

[Chris Faylor]

On Mon, May 29, 2000 at 07:46:57PM +0300, Ville Herva wrote:
>On Mon, May 29, 2000 at 11:59:11AM +0300, you [Ville Herva] claimed:
>> I must be doing something wrong here. Basicly, it seems like sed would
>> always intepret its input as binary.
>> 
>> I'm trying to run zsh-3.1.6 configure script. All is fine until configure
>> does something like
>> 
>> cat > c.sed << EOF
>> s/ristiina/parkano/g
>> s/oulu/turku/g
>> EOF
>> sed -n -f c.sed < naantali > hollola
>> 
>> Now, c.sed contains dos newlines (0x0d), and sed barfs:
>> 
>> sed: file conftest.frag line 1: Unknown option to 's'
>> 
>> (option 'g^M'...)
>> 
>> my mounts are textmode, but the situation did not change with binary
>> either. The configure script (the cat > ... and the following lines in
>> particular) has no ^M's. CYGWIN=binmode is _not_ set.
>> 
>> cygwin1.ddl snapshot 15052000, newest sed from net.
>
>This is a problem with sed. I tried to recompile sed but its configure
>script failed the same way! How do you people get configure scripts run? I
>couldn't even figure out any bearable work-around.
>
>Sed uses mmap to read executed file:
>
>sed.c::main():
>        case 'f':
>          the_program = compile_file(the_program, optarg);
>          break;
>
>compile.c:.compile_file():
>  if (cmdfile[0] != '-' || cmdfile[1] != '\0')
>    prog.file = ck_fopen(cmdfile, "r");
>  if (map_file(prog.file, &prog.base, &len))
>  
>  
>util.c::cd_fopen():
>  if ( ! (fp = fopen(name, mode)) )
>  
>  
>sed.c::map_file():
>      nbase = VCAST(VOID *)mmap(NULL, CAST(size_t)s.st_size, PROT_READ,
>                                MAP_PRIVATE, fileno(fp), CAST(off_t)0);
>                                
>
>I assume CR/LF translation is not done for mmap? 
>
>Am I missing something obvious?

Nope.  Patches are gratefully accepted.

Until then, endure the horror of removing \r endings from your sed files.

cgf

--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

Re: sed and configure scripts

[Ville Herva]

On Mon, May 29, 2000 at 12:56:16PM -0400, you [Chris Faylor] claimed:
> 
> Nope.  Patches are gratefully accepted.

Patching doesn't seem to be impossible even for me -- it seems the mmap
could just be disabled so that sed falls back to ordinary file handling.
The problem however is that I can't compile sed, because sed does not
work... Bummer.

> Until then, endure the horror of removing \r endings from your sed files.

Ugh.


-- v --

v@iki.fi

--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com

Re: sed and configure scripts

[Ville Herva]

> Nope.  Patches are gratefully accepted.

Yep. Moving from 20000515 snapshot to 20000525 snapshot fixed the problem.
(0526 and 0527 snapshots were broken.)
 

-- v --

v@iki.fi

--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com