Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

Codesighs Reports

2 views
Skip to first unread message

Shawn Wilsher

unread,
Oct 26, 2010, 10:25:35 AM10/26/10
to dev-tree-management
Hey all,

Do we just not care about this codesighs regressions that have been
coming through? Nobody ever comments on them...

Cheers,

Shawn

Mike Beltzner

unread,
Oct 26, 2010, 10:26:47 AM10/26/10
to Shawn Wilsher, dev-tree-management
On 2010-10-26, at 10:25 AM, Shawn Wilsher wrote:

> Do we just not care about this codesighs regressions that have been coming through? Nobody ever comments on them...

I certainly don't know how to reason about them. Are they our compiled codesize or active codesize? Does increasing test coverage increase our codesize?

cheers,
mike

L. David Baron

unread,
Oct 26, 2010, 1:00:13 PM10/26/10
to Shawn Wilsher, dev-tree-management
On Tuesday 2010-10-26 07:25 -0700, Shawn Wilsher wrote:
> Do we just not care about this codesighs regressions that have been
> coming through? Nobody ever comments on them...

I don't think we care about a 2K codesize regression if somebody
added a bunch of new code. That's expected.

We do care about unexpected codesize regressions (e.g., 1 meg change
in codesize because of change to optimization flags or a change to
something in nsString.h or nsCOMPtr.h).

It might make sense so we don't get notified on any codesize changes
less than, say 10K or 20K.

-David

--
L. David Baron http://dbaron.org/
Mozilla Corporation http://www.mozilla.com/

Mike Shaver

unread,
Oct 26, 2010, 1:02:25 PM10/26/10
to L. David Baron, Shawn Wilsher, dev-tree-management
On Tue, Oct 26, 2010 at 10:00 AM, L. David Baron <dba...@dbaron.org> wrote:
> I don't think we care about a 2K codesize regression if somebody
> added a bunch of new code.  That's expected.
>
> We do care about unexpected codesize regressions (e.g., 1 meg change
> in codesize because of change to optimization flags or a change to
> something in nsString.h or nsCOMPtr.h).
>
> It might make sense so we don't get notified on any codesize changes
> less than, say 10K or 20K.

Exactly so.

Mike

Ehsan Akhgari

unread,
Oct 27, 2010, 12:28:06 AM10/27/10
to Mike Shaver, Shawn Wilsher, L. David Baron, dev-tree-management

Filed bug 607552 so that we don't forget about this.

Ehsan

0 new messages