Hi Caolán, On Thursday, 2011-09-01 10:03:40 +0100, Caolán McNamara wrote:
On Wed, 2011-08-31 at 23:54 -0600, Tor Lillqvist wrote:The attached patch is an attempt to bring some sanity to the situation:It seems that you have investigate this quite deeply and I would love to commit and push your patch. However, I am a bit scared. Could somebody else who actually understands the issues involved have a look?I think we're all too chicken to man-up and look at this one :-)
/me ducks
I guess the major fear isn't that your code is wrong, it's probably right, it's what stuff relying on the old way of doing things will change. I've been planning, but delayed on tracking down a mystery 2 point offset :-), to put together a "did my layout change" regression test which would have been perfect to test this.
This indeed may be a problem, users expect documents to render identical, even if it was wrong before.
caolan->tor: I suggest bunging this in for now, but putting the old algorithm behind an getenv(SAL_OLD_METRICS) or something for easy side-by-side testing ?
Good idea. I'd just rename that to VCL_OLD_FONT_METRICS for clarity. We could also make use of the convwatch tests. We'll also need comparisons with MS applications for the same document.
I'd be interested in knowing what effects the change has on typical fonts as opposed to the edge cases, i.e. does text rendered in Liberation Serif, Sans, Monotype stay basically the same.
Apparently not. Eike -- PGP/OpenPGP/GnuPG encrypted mail preferred in all private communication. Key ID: 0x293C05FD - 997A 4C60 CE41 0149 0DB3 9E96 2F1A D073 293C 05FD
Attachment:
signature.asc
Description: Digital signature