On 07/27/2017 08:25 PM, Philippe Mathieu-Daudé wrote:
> Hi, following Cleber Rosa example I cleaned more invalid references.
>
> Eric said this can wait 2.11, however these patches don't change any code
> generated, I think the 2.10 users deserve an up-to-date doc :p
I think what I said was: this doesn't change code and doc updates are
find during freeze, so this is a GOOD candidate for 2.10; but if it
misses 2.10, slipping to 2.11 doesn't hurt. (If I didn't say that, it's
what I should have done). But you caught on to my intention, even if in
the process it sounds like I was asking you to wait.
>
> I used the following command (and consider include it in some CI test job):
>
> $ git grep docs/ \
> | sed -ne "s/.* \(docs[^ :)}\"\']*\).*/\1/p" \
> | sed -e 's/\(.*\)\.$/\1/p' | sort -u | while read p;do
> ls -ld $p 1>/dev/null
Wait. What? 'ls -ld' with stdout sent to /dev/null? (I never cease to
be amazed at the random shell constructs that people come up with). I
guess you are just trying to check for file existence, by going off of
the stderr droppings from ls? 'ls -d' is less effort than 'ls -ld'; and
if all you care about is stderr messages, spawning one ls per filename
is much slower than just doing it wholesale via xargs to minimize ls calls:
git grep ... | sort -u | xargs ls -d >/dev/null
But then again, efficiency in generating the list of problems isn't a
bottleneck.
--
Eric Blake, Principal Software Engineer
Red Hat, Inc. +1-919-301-3266
Virtualization: qemu.org | libvirt.org