On Wed, Dec 11, 2019 at 12:52:20PM -0600, Eric Blake wrote:
On 12/11/19 12:45 PM, Daniel P. Berrangé wrote:
> This avoids leaving a zero length or partially generated output
> file on errors.
>
> Signed-off-by: Daniel P. Berrangé <berrange(a)redhat.com>
> ---
> docs/Makefile.am | 10 +++++-----
> 1 file changed, 5 insertions(+), 5 deletions(-)
>
> diff --git a/docs/Makefile.am b/docs/Makefile.am
> index eb8de80b9c..9a1f7a6117 100644
> --- a/docs/Makefile.am
> +++ b/docs/Makefile.am
> @@ -259,21 +259,21 @@ man8_MANS = $(manpages8_rst:%.rst=%.8)
> grep -v '^\.\. contents::' < $< | \
> sed -e 's|SYSCONFDIR|$(sysconfdir)|g' \
> -e 's|RUNSTATEDIR|$(runstatedir)|g' | \
> - $(RST2MAN) > $@
> + $(RST2MAN) > $@ || { rm $@ && exit 1; }
But still allows a truncated view of the file if another process accesses
the file while RST2MAN is still running. Even better is to generate output
to a temp file then atomically mv it into place, so that no concurrent
process can ever see an incomplete file.
Is that a problem that actually impacts negatively in real world for builds?
In various places in the make rules, we've either done the "|| rm"
approach,
or the temp file and rename approach. Personally I find the recipes using
the "|| rm" approach are more maintainable/readable.
Regards,
Daniel
--
|:
https://berrange.com -o-
https://www.flickr.com/photos/dberrange :|
|:
https://libvirt.org -o-
https://fstop138.berrange.com :|
|:
https://entangle-photo.org -o-
https://www.instagram.com/dberrange :|