Ports of lib9, libbio and libmach to Windows.

Ports of lib9, libbio and libmach to Windows.

[rsc, 2009-11-24 23:11:55]

Thanks for doing this.

These seem reasonable.  As you might have guessed
from the copyright notices, a lot of the support code
here was inherited from another project, and a lot of 
it is not necessary here.  So rather than tweak things to
keep them compiling, I am all for throwing away things
that aren't necessary right now: I don't see our system
requirements growing on the C side.

I'd like to keep the number of #ifdefs to a minimum.
Some are unavoidable, like the one in p9times, but
many can be consolidated into the single __MINGW__
check in libc.h.  I've marked those below.

http://codereview.appspot.com/157159/diff/1018/24
File src/lib9/create.c (right):

http://codereview.appspot.com/157159/diff/1018/24#newcode56
src/lib9/create.c:56: if(mkdir(path) < 0)
I'd rather this was in libc.h in the #ifdef __MINGW32__

#define mkdir(path, perm) mkdir(path)

http://codereview.appspot.com/157159/diff/1018/24#newcode81
src/lib9/create.c:81: #ifdef __MINGW32__
Same here - if not defined __MINGW32__ then 
#define O_BINARY 0
in libc.h

http://codereview.appspot.com/157159/diff/1018/24#newcode88
src/lib9/create.c:88: #ifndef __MINGW32__
You can throw this code away, and take
OEXCL and ORCLOSE out of libc.h.
No code uses it, so it's not worth the #ifdef.

http://codereview.appspot.com/157159/diff/1018/27
File src/lib9/exec.c (right):

http://codereview.appspot.com/157159/diff/1018/27#newcode32
src/lib9/exec.c:32: return execvp(prog, (const char**)argv);
Another for the libc.h file

#define execvp(prog, argv) execvp(prog, (const char**)(argv))

Believe it or not, it should work.

http://codereview.appspot.com/157159/diff/1018/28
File src/lib9/open.c (right):

http://codereview.appspot.com/157159/diff/1018/28#newcode55
src/lib9/open.c:55: #ifndef __MINGW32__
Can throw away ONONBLOCK from the libc.h file
and then this code too.

http://codereview.appspot.com/157159/diff/1018/28#newcode70
src/lib9/open.c:70: umode |= O_BINARY;
Once O_BINARY is in libc.h won't need the #ifdef.

http://codereview.appspot.com/157159/diff/1018/28#newcode74
src/lib9/open.c:74: #ifndef __MINGW32__
Same comment - can throw this away.

http://codereview.appspot.com/157159/diff/1018/29
File src/lib9/pipe.c (right):

http://codereview.appspot.com/157159/diff/1018/29#newcode39
src/lib9/pipe.c:39: p9pipe(int fd[2])
Should probably throw this file away too
and get rid of the #define about pipe in libc.h.
The standard pipe is fine.

If you need to, you can put

#define pipe(fd) _pipe(fd, 512, O_BINARY)

in the mingw section of libc.h.

http://codereview.appspot.com/157159/diff/1018/33
File src/libbio/bseek.c (right):

http://codereview.appspot.com/157159/diff/1018/33#newcode81
src/libbio/bseek.c:81: #ifdef __MINGW32__
Instead, in libc.h

#define lseek(fd, n, base) _lseeki64(fd, n, base)

[hector, 2009-11-25 00:31:21]

Thanks for taking the time to review the changes.  I've taken onboard your
comments and have modified the patch accordingly.

On 2009/11/24 23:11:55, rsc wrote:
> Thanks for doing this.
> 
> These seem reasonable.  As you might have guessed
> from the copyright notices, a lot of the support code
> here was inherited from another project, and a lot of 
> it is not necessary here.  So rather than tweak things to
> keep them compiling, I am all for throwing away things
> that aren't necessary right now: I don't see our system
> requirements growing on the C side.
> 
> I'd like to keep the number of #ifdefs to a minimum.
> Some are unavoidable, like the one in p9times, but
> many can be consolidated into the single __MINGW__
> check in libc.h.  I've marked those below.
> 
> http://codereview.appspot.com/157159/diff/1018/24
> File src/lib9/create.c (right):
> 
> http://codereview.appspot.com/157159/diff/1018/24#newcode56
> src/lib9/create.c:56: if(mkdir(path) < 0)
> I'd rather this was in libc.h in the #ifdef __MINGW32__
> 
> #define mkdir(path, perm) mkdir(path)
> 
> http://codereview.appspot.com/157159/diff/1018/24#newcode81
> src/lib9/create.c:81: #ifdef __MINGW32__
> Same here - if not defined __MINGW32__ then 
> #define O_BINARY 0
> in libc.h
> 
> http://codereview.appspot.com/157159/diff/1018/24#newcode88
> src/lib9/create.c:88: #ifndef __MINGW32__
> You can throw this code away, and take
> OEXCL and ORCLOSE out of libc.h.
> No code uses it, so it's not worth the #ifdef.
> 
> http://codereview.appspot.com/157159/diff/1018/27
> File src/lib9/exec.c (right):
> 
> http://codereview.appspot.com/157159/diff/1018/27#newcode32
> src/lib9/exec.c:32: return execvp(prog, (const char**)argv);
> Another for the libc.h file
> 
> #define execvp(prog, argv) execvp(prog, (const char**)(argv))
> 
> Believe it or not, it should work.
> 
> http://codereview.appspot.com/157159/diff/1018/28
> File src/lib9/open.c (right):
> 
> http://codereview.appspot.com/157159/diff/1018/28#newcode55
> src/lib9/open.c:55: #ifndef __MINGW32__
> Can throw away ONONBLOCK from the libc.h file
> and then this code too.
> 
> http://codereview.appspot.com/157159/diff/1018/28#newcode70
> src/lib9/open.c:70: umode |= O_BINARY;
> Once O_BINARY is in libc.h won't need the #ifdef.
> 
> http://codereview.appspot.com/157159/diff/1018/28#newcode74
> src/lib9/open.c:74: #ifndef __MINGW32__
> Same comment - can throw this away.
> 
> http://codereview.appspot.com/157159/diff/1018/29
> File src/lib9/pipe.c (right):
> 
> http://codereview.appspot.com/157159/diff/1018/29#newcode39
> src/lib9/pipe.c:39: p9pipe(int fd[2])
> Should probably throw this file away too
> and get rid of the #define about pipe in libc.h.
> The standard pipe is fine.
> 
> If you need to, you can put
> 
> #define pipe(fd) _pipe(fd, 512, O_BINARY)
> 
> in the mingw section of libc.h.
> 
> http://codereview.appspot.com/157159/diff/1018/33
> File src/libbio/bseek.c (right):
> 
> http://codereview.appspot.com/157159/diff/1018/33#newcode81
> src/libbio/bseek.c:81: #ifdef __MINGW32__
> Instead, in libc.h
> 
> #define lseek(fd, n, base) _lseeki64(fd, n, base)

[rsc, 2009-11-25 00:57:21]

This looks great.  Could you please 
complete one of the license agreements
as described at http://golang.org/doc/contribute.html#copyright 
and then let me know that you have?

Don't worry about sending a CL with AUTHORS and CONTRIBUTORS changes - I'll make
the change once
I've verified that the agreement worked.

This is a one-time thing - future CLs won't require
redoing this.

Thanks.

[hector, 2009-11-25 09:53:44]

Done.

2009/11/25  <rsc@golang.org>:
> This looks great.  Could you please
> complete one of the license agreements
> as described at http://golang.org/doc/contribute.html#copyright
> and then let me know that you have?
>
> Don't worry about sending a CL with AUTHORS and CONTRIBUTORS changes -
> I'll make the change once
> I've verified that the agreement worked.
>
> This is a one-time thing - future CLs won't require
> redoing this.
>
> Thanks.
>
> http://codereview.appspot.com/157159
>

[rsc, 2009-11-30 19:46:44]

i'm about to submit this (you'll see another mail).
the make.bash change didn't
apply quite cleanly; the easiest
way to make your sync work
cleanly is to revert make.bash first:

cd $GOROOT
hg revert src/make.bash
hg sync

[rsc, 2009-11-30 19:53:14]

*** Submitted as http://code.google.com/p/go/source/detail?r=5c440ffb5ac8 ***

Ports of lib9, libbio and libmach to Windows.

R=rsc
http://codereview.appspot.com/157159

Committer: Russ Cox <rsc@golang.org>