--===============5932222825907839898== Content-Type: multipart/alternative; boundary="===============2465256475904870495==" --===============2465256475904870495== MIME-Version: 1.0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://git.reviewboard.kde.org/r/127632/#review94759 ----------------------------------------------------------- Ship it! i think adhering spec here is not very important. we can adopt this change now, mention this somewhere in docs (print a qWarning saying that the checking order has changed?), and introduce X-KDE-Extensions later (which would only require a slght adjustment of iconPathByName() method). - Nick Shaforostoff On April 11, 2016, 11:17 p.m., Aleix Pol Gonzalez wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://git.reviewboard.kde.org/r/127632/ > ----------------------------------------------------------- > > (Updated April 11, 2016, 11:17 p.m.) > > > Review request for KDE Frameworks and Christoph Feck. > > > Repository: kiconthemes > > > Description > ------- > > Usually themes will have 1 kind of extension (2 tops) it doesn't make sense to bindly use an extensions vector that is statically defined. > > Prioritizes the extensions that work, so it will find the icons sooner. > > > Diffs > ----- > > src/kiconloader.cpp 37528ad > src/kicontheme.h 3190665 > src/kicontheme.cpp e7e003b > > Diff: https://git.reviewboard.kde.org/r/127632/diff/ > > > Testing > ------- > > Builds and tests still pass. > Also reducess faulty disk accesses to its 45%. Note that by this metric, it's almost as good as the RR #127236, but with a much smaller memory impact, since we're not caching all the icon names (which was specially bad as it was per-process). > ``` > $ strace kwrite |& grep ENOENT | wc -l > 4699 > $ strace kwrite |& grep ENOENT | wc -l > 2119 > ``` > > > Thanks, > > Aleix Pol Gonzalez > > --===============2465256475904870495== MIME-Version: 1.0 Content-Type: text/html; charset="utf-8" Content-Transfer-Encoding: 7bit

This is an automatically generated e-mail. To reply, visit: https://git.reviewboard.kde.org/r/127632/

Ship it!

i think adhering spec here is not very important.

we can adopt this change now, mention this somewhere in docs (print a qWarning saying that the checking order has changed?), and introduce X-KDE-Extensions later (which would only require a slght adjustment of iconPathByName() method).

- Nick Shaforostoff

On April 11th, 2016, 11:17 p.m. UTC, Aleix Pol Gonzalez wrote:

Review request for KDE Frameworks and Christoph Feck.

By Aleix Pol Gonzalez.

Updated April 11, 2016, 11:17 p.m.

Repository: kiconthemes

Description

Usually themes will have 1 kind of extension (2 tops) it doesn't make sense to bindly use an extensions vector that is statically defined.

Prioritizes the extensions that work, so it will find the icons sooner.

Testing

Builds and tests still pass. Also reducess faulty disk accesses to its 45%. Note that by this metric, it's almost as good as the RR #127236, but with a much smaller memory impact, since we're not caching all the icon names (which was specially bad as it was per-process).

$ strace kwrite |& grep ENOENT | wc -l
4699
$ strace kwrite |& grep ENOENT | wc -l
2119

Diffs

src/kiconloader.cpp (37528ad)
src/kicontheme.h (3190665)
src/kicontheme.cpp (e7e003b)

--===============2465256475904870495==-- --===============5932222825907839898== Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: base64 Content-Disposition: inline X19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX18KS2RlLWZyYW1l d29ya3MtZGV2ZWwgbWFpbGluZyBsaXN0CktkZS1mcmFtZXdvcmtzLWRldmVsQGtkZS5vcmcKaHR0 cHM6Ly9tYWlsLmtkZS5vcmcvbWFpbG1hbi9saXN0aW5mby9rZGUtZnJhbWV3b3Jrcy1kZXZlbAo= --===============5932222825907839898==--