Re: [Gnu-arch-users] corrupted binary files

[cduffy]

On Fri, Apr 02, 2004 at 01:10:15PM +1000, Brian May wrote:
> >>>>> "Charles" == Charles Duffy <address@hidden> writes:
>
>     Charles> On Thu, 2004-04-01 at 20:32, Brian May wrote:
>     >> These binary files were imported using cscvs (not sure if that
>     >> makes any difference...).
>
>     Charles> It very well might. See if cvs -q update -p -r${REVISION}
>     Charles> ${FILENAME} > ${FILENAME} gives correct or incorrect
>     Charles> output.
>
> Interesting... Same corrupted output.
>
> However, the file in the checked-out CVS tree is correct.
>
> Could it be that the -p parameter is doing something weird?

Presumably. -p requests that a checked-out file be piped to stdout; cscvs
routes from that pipe to the destination (which is frequently *not* inside a
CVS repository, which is why a plain "cvs update" won't typically Do The Right
Thing; I don't want to modify the repository cscvs is being run from).

An alternate mechanism avoiding the -p has been proposed (to preserve +x
permissions); I didn't merge it on account of ugliness (which is kind of silly
now considering some of the code that's made it in). Given your report, I'll be
reconsidering the patch -- or some alternate solution.

> Does cscvs use the -p parameter?

The command I gave you is exactly what cscvs uses internally.