Re: [PATCH] Adding systemtap probe points in pthread library (slightly revised again)

[Bert Wesarg <bert dot wesarg at googlemail dot com>]

On Wed, Feb 9, 2011 at 19:34, Rayson Ho <rho@redhat.com> wrote:
> (Didn't include the diff for DESIGN-systemtap-probes.txt in the email I
> sent 1 min ago, resending...)
>
> On Wed, 2011-02-09 at 19:00 +0100, Bert Wesarg wrote:
>> IMHO, this is a typo. Because for 'read' you have
>> 'rdlock_acquire_read' and 'rdlock_entry'. So judging from this
>> pattern, I would expect, that this probe would be
>> 'wrlock_acquire_write', like 'wrlock_entry'. Also note, that in the
>> 'acquire' probe names are redundancies for 'read' and 'write' (after
>> fixing the typo), while in the 'entry' names not.
>
> Thanks Bert for providing comments since the beginning of the release of
> this patch. I think changing wrlock_acquire_write makes sense -
> originally I wanted to use "rw" to indicate it's a read-write lock, and
> use "wr" to indicate it's "write". But in this case it does not make
> things consistent.
>
> However, I think it makes sense to have the 'acquire' probe names - but
> if you really think there are redundancies, can you explain it in a bit
> more detail??

Sure. You have now 'wrlock_acquire_write'. The prefix 'wr' tells the
reader that its about a write lock, so the suffix 'write' is
redundant, isn't ist? You haven't this redundancy in the
'wrlock_entry' case, else it would have the name 'wrlock_entry_write'.

Bert