BSD port: USE_NSS and other macros instead of OS_LINUX where applicable

BSD port: USE_NSS and other macros instead of OS_LINUX where applicable


Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=37492

[Evan Martin, 2010-01-29 01:21:38]

http://codereview.chromium.org/556068/diff/1/9
File net/base/host_resolver_proc.cc (right):

http://codereview.chromium.org/556068/diff/1/9#newcode81
net/base/host_resolver_proc.cc:81: // On Linux changes to /etc/resolv.conf can
go unnoticed thus resulting in
Maybe update this "Linux" here.  Your call.
(Don't worry if the resulting diff is large when you reflow the comment -- I
don't care too much.)

http://codereview.chromium.org/556068/diff/1/9#newcode209
net/base/host_resolver_proc.cc:209: #if defined(OS_LINUX) || defined(OS_FREEBSD)
This FREEBSD-only bit contradicts the above part (where I suggested you edit the
comment).

[Peter Valchev, 2010-01-29 02:06:55]

http://codereview.chromium.org/556068/diff/1/9
File net/base/host_resolver_proc.cc (right):

http://codereview.chromium.org/556068/diff/1/9#newcode81
net/base/host_resolver_proc.cc:81: // On Linux changes to /etc/resolv.conf can
go unnoticed thus resulting in
On 2010/01/29 01:21:38, Evan Martin wrote:
> Maybe update this "Linux" here.  Your call.
> (Don't worry if the resulting diff is large when you reflow the comment -- I
> don't care too much.)

Done

http://codereview.chromium.org/556068/diff/1/9#newcode209
net/base/host_resolver_proc.cc:209: #if defined(OS_LINUX) || defined(OS_FREEBSD)
On 2010/01/29 01:21:38, Evan Martin wrote:
> This FREEBSD-only bit contradicts the above part (where I suggested you edit
the
> comment).

Right, thanks for catching this... OpenBSD is affected by this and needs this
hack just as much as Linux/FreeBSD but there is no support for
res_ninit/res_nclose. I wonder how they differ from res_init/res_close? There is
no manpage for res_ninit on linux :(

[Evan Martin, 2010-01-29 02:14:02]

http://codereview.chromium.org/556068/diff/1/9
File net/base/host_resolver_proc.cc (right):

http://codereview.chromium.org/556068/diff/1/9#newcode209
net/base/host_resolver_proc.cc:209: #if defined(OS_LINUX) || defined(OS_FREEBSD)
On 2010/01/29 02:06:55, Peter Valchev wrote:
> On 2010/01/29 01:21:38, Evan Martin wrote:
> > This FREEBSD-only bit contradicts the above part (where I suggested you edit
> the
> > comment).
> 
> Right, thanks for catching this... OpenBSD is affected by this and needs this
> hack just as much as Linux/FreeBSD but there is no support for
> res_ninit/res_nclose. I wonder how they differ from res_init/res_close? There
is
> no manpage for res_ninit on linux :(

If you don't want to worry about it for now, you can add an #elif
defined(OS_OPENBSD) below with NOTIMPLEMENTED() in its  body.  It will complain
to the console every time Chrome resolves anything, though...

[Peter Valchev, 2010-01-29 02:27:14]

On 2010/01/29 02:14:02, Evan Martin wrote:
> http://codereview.chromium.org/556068/diff/1/9
> File net/base/host_resolver_proc.cc (right):
> 
> http://codereview.chromium.org/556068/diff/1/9#newcode209
> net/base/host_resolver_proc.cc:209: #if defined(OS_LINUX) ||
defined(OS_FREEBSD)
> On 2010/01/29 02:06:55, Peter Valchev wrote:
> > On 2010/01/29 01:21:38, Evan Martin wrote:
> > > This FREEBSD-only bit contradicts the above part (where I suggested you
edit
> > the
> > > comment).
> > 
> > Right, thanks for catching this... OpenBSD is affected by this and needs
this
> > hack just as much as Linux/FreeBSD but there is no support for
> > res_ninit/res_nclose. I wonder how they differ from res_init/res_close?
There
> is
> > no manpage for res_ninit on linux :(
> 
> If you don't want to worry about it for now, you can add an #elif
> defined(OS_OPENBSD) below with NOTIMPLEMENTED() in its  body.  It will
complain
> to the console every time Chrome resolves anything, though...

Actually I just looked at what Mozilla does and since res_ninit is not
available, it just disables this feature on OpenBSD. I did the same here and
added a comment saying it is not supported.

I guess res_ninit is just a thread-safe version of res_init. I'll add this on my
todo list to add to OpenBSD and then we can revisit this, but for now I think
this is best. Uploaded new CL.

[Evan Martin, 2010-01-29 02:38:09]

LGTM, let me know when you commit it and I'll file a bug about the res_ninit
thing so we don't forget.

[Peter Valchev, 2010-01-29 03:36:59]

On 2010/01/29 02:38:09, Evan Martin wrote:
> LGTM, let me know when you commit it and I'll file a bug about the res_ninit
> thing so we don't forget.

Committed

[Evan Martin, 2010-01-29 07:22:52]

OK, filed http://code.google.com/p/chromium/issues/detail?id=33481

[wtc, 2010-01-30 00:09:16]

LGTM.  Sorry about the delay.

http://codereview.chromium.org/556068/diff/16/18
File net/socket/client_socket.h (right):

http://codereview.chromium.org/556068/diff/16/18#newcode12
net/socket/client_socket.h:12: #include <sys/types.h>
<sys/types.h> should not need to be included for
<sys/socket.h> in new Unix systems:

http://www.opengroup.org/onlinepubs/009695399/functions/socket.html

But I know <sys/types.h> used to be required.

[Peter Valchev, 2010-02-01 19:12:18]

http://codereview.chromium.org/556068/diff/16/18
File net/socket/client_socket.h (right):

http://codereview.chromium.org/556068/diff/16/18#newcode12
net/socket/client_socket.h:12: #include <sys/types.h>
On 2010/01/30 00:09:16, wtc wrote:
> <sys/types.h> should not need to be included for
> <sys/socket.h> in new Unix systems:
> 
> http://www.opengroup.org/onlinepubs/009695399/functions/socket.html
> 
> But I know <sys/types.h> used to be required.

It is still required both on OpenBSD and FreeBSD:

http://www.openbsd.org/cgi-bin/man.cgi?query=socket&apropos=0&sektion=0&manpa...

http://www.freebsd.org/cgi/man.cgi?query=socket&apropos=0&sektion=0&manpath=F...