668315 - SMILIntegerType::Interpolate should use NS_lround instead of re-implementing rounding

Status: RESOLVED WONTFIX

(Core :: SVG, defect)

Description

[Daniel Holbert [:dholbert]]

SMILIntegerType::Interpolate has some manual round-to-nearest int code that I think can be replaced with a call to NS_lround().

Comment 1

[Daniel Holbert [:dholbert]]

Specifically:
86   // When currentVal is exactly midway between its two nearest integers, we
87   // jump to the "next" integer to provide simple, easy to remember and
88   // consistent behaviour (from the SMIL author's point of view).
89 
90   if (startVal < endVal) {
91     aResult.mU.mInt = int64_t(floor(currentVal + 0.5)); // round mid up
92   } else {
93     aResult.mU.mInt = int64_t(ceil(currentVal - 0.5)); // round mid down
94   }
http://mxr.mozilla.org/mozilla-central/source/content/smil/SMILIntegerType.cpp#86

Though note that NS_lround returns a 32-bit int, not a 64-bit int, so it may not be appropriate here after all.

Comment 2

[Robert Longson [:longsonr]]

Attachment: Bug 668315 - Use std:llround rather than re-implementing rounding r=dholbert

Comment 3

[Robert Longson [:longsonr]]

https://treeherder.mozilla.org/jobs?repo=try&revision=297121d09556e9ff223f1d8be96470964840d0e3

Comment 4

[Robert Longson [:longsonr]]

The standard library only rounds in one direction so it's not going to be simpler than the code we have now.