bsdiff: if file is /dev/fd/*, don't open/close it. Just use the fd.

bsdiff: if file is /dev/fd/*, don't open/close it. Just use the fd.

The updater, in an attempt to keep bspatch from open/closing the raw
block device frequently, passes a file descriptor and a path of
/dev/fd/(some number) as the path. This patch causes bspatch to
recognize that magic file path and not call open/close.

This dramatically reduces resource usage on client machines performing
delta updates.

BUG=7636
TEST=Installed delta on device

Committed: http://chrome-svn/viewvc/chromeos?view=rev&revision=a13e3cb

[petkov, 2010-10-12 21:53:23]

LGTM

http://codereview.chromium.org/3698005/diff/1/2
File dev-util/bsdiff/files/4.3_bspatch-support-input-output-positioning.patch
(right):

http://codereview.chromium.org/3698005/diff/1/2#newcode189
dev-util/bsdiff/files/4.3_bspatch-support-input-output-positioning.patch:189:
+		sscanf(filename + strlen(kFdFilePrefix), "%d", &fd);
given that you're not detecting errors, why not use atoi?

[adlr, 2010-10-12 23:07:45]

On Tue, Oct 12, 2010 at 2:53 PM, <petkov@chromium.org> wrote:

> LGTM
>
>
>
> http://codereview.chromium.org/3698005/diff/1/2
> File
>
> dev-util/bsdiff/files/4.3_bspatch-support-input-output-positioning.patch
> (right):
>
> http://codereview.chromium.org/3698005/diff/1/2#newcode189
>
> dev-util/bsdiff/files/4.3_bspatch-support-input-output-positioning.patch:189:
> +               sscanf(filename + strlen(kFdFilePrefix), "%d", &fd);
> given that you're not detecting errors, why not use atoi?


I'm working on the assumption that if there's a failure, fd will remain -1,
but w/ atoi() I'm not sure what the error return would be.

-andrew


>
>
> http://codereview.chromium.org/3698005/show
>

[petkov, 2010-10-12 23:10:37]

On 2010/10/12 23:07:45, adlr wrote:
> On Tue, Oct 12, 2010 at 2:53 PM, <mailto:petkov@chromium.org> wrote:
> 
> > LGTM
> >
> >
> >
> > http://codereview.chromium.org/3698005/diff/1/2
> > File
> >
> > dev-util/bsdiff/files/4.3_bspatch-support-input-output-positioning.patch
> > (right):
> >
> > http://codereview.chromium.org/3698005/diff/1/2#newcode189
> >
> >
dev-util/bsdiff/files/4.3_bspatch-support-input-output-positioning.patch:189:
> > +               sscanf(filename + strlen(kFdFilePrefix), "%d", &fd);
> > given that you're not detecting errors, why not use atoi?
> 
> 
> I'm working on the assumption that if there's a failure, fd will remain -1,
> but w/ atoi() I'm not sure what the error return would be.

A true... still LGTM.

Thanks,

Darin

> 
> -andrew
> 
> 
> >
> >
> > http://codereview.chromium.org/3698005/show
> >
>