Subject: Re: [Boost-build] Problem with Install Rule
From: Vladimir Prus (ghost_at_[hidden])
Date: 2008-10-25 13:48:08
On Saturday 25 October 2008 21:39:59 Steven Watanabe wrote:
> Vladimir Prus wrote:
> > thanks for addressing this! The fact that copy fails to change timestamp
> > never occurred to me.
> > But can you explain how the patch works? Ideally, as a quick comment in the
> > patch itself. It appears that /B and /Y options are unrelated to timestamp
> > issues, according to my reading -- one forces binary mode, and another
> > disables prompting if destination exists (but I did not got any error anyway).
> > Does
> > copy a + nonexistent_file dest
> > actually changes the timestamp, for some reason?
> a) The command
> copy a + b dest
> concatenates the files a and b and writes the result to dest.
> This causes the timestamp of dest to be set to the current time.
> b) the error checking of copy does not seem to be very good,
> so it doesn't issue an error if the second file doesn't exist.
> I'm not sure how portable this is.
> c) the normal default is to copy files in binary mode. However,
> copying multiple files to a single destination causes the default
> to be ascii mode. This can truncate files. That's why I had to add /b.
> d) /Y is not needed, as it is the default unless
> copy is run interactively.
Thanks for explaining. If you could mention this in the code,
it would be great.
Boost-Build list run by bdawes at acm.org, david.abrahams at rcn.com, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk