/irc-logs / w3c / #html-wg / 2007-12-02 / end

Options:

  1. # Session Start: Sun Dec 02 00:00:00 2007
  2. # Session Ident: #html-wg
  3. # [00:00] * Quits: Lachy (Lachlan@80.202.68.140) (Quit: This computer has gone to sleep)
  4. # [00:03] * Joins: jgraham_ (james@81.86.218.70)
  5. # [00:16] * Joins: Lachy (Lachlan@84.215.41.149)
  6. # [00:33] <mjs> Lachy: am now
  7. # [00:34] <Lachy> mjs, in WebKit's <video> implementation, why does the image loaded by the poster attribute only display until the video has loaded? Shouldn't it display until the video starts playing?
  8. # [00:35] <mjs> Lachy: could be a bug, but I thought it did the later (wait until video first plays)
  9. # [00:36] <Lachy> nope
  10. # [00:38] * Quits: aroben (aroben@67.160.250.192) (Connection reset by peer)
  11. # [00:52] <mjs> Lachy: can you file it (<http://bugs.webkit.org>) or should I?
  12. # [00:53] <Lachy> you can
  13. # [01:02] <mjs> Lachy: filed http://bugs.webkit.org/show_bug.cgi?id=16226
  14. # [01:10] * Quits: gsnedders (gsnedders@86.145.188.131) (Quit: gsnedders)
  15. # [01:10] <Lachy> mjs, you'll need to change that link in the bug report, since that's in my temp directory and will be deleted shortly
  16. # [01:11] <Lachy> (also, that directory was supposed to be hidden, which is why I sent you the link privately)
  17. # [01:11] <mjs> Lachy: ah, sorry
  18. # [01:11] <Lachy> I'll make a proper test case, give me a few minutes
  19. # [01:11] <mjs> Lachy: I'll delete the URL
  20. # [01:12] <Philip> Lachy: If you say something was "supposed to be hidden" on a public channel, you know everybody's going to rush to look at it and poke around to see what else they can find in that directory ;-)
  21. # [01:13] <Lachy> it's gone now!
  22. # [01:14] <mjs> Lachy: sorry about that, just bug filing habite
  23. # [01:14] <mjs> *habit
  24. # [01:28] * Joins: myakura (myakura@122.29.186.5)
  25. # [01:32] <Lachy> mjs, http://lachy.id.au/dev/markup/tests/html5/video/007 and http://lachy.id.au/dev/markup/tests/html5/video/008
  26. # [01:40] <Philip> Oh, I just remembered it's probably a bad idea to try to W3Cise the canvas tests now, since the WG might decide to exclude canvas and force all the useful work to happen somewhere else
  27. # [01:42] <mjs> Philip: and here we witness the amazing power of FUD in stopping progress (though I think that is probably a wise choice on your part)
  28. # [01:43] <Philip> mjs: It's only stopping progress in the W3Cising direction, which leaves more time for progress in e.g. the integrating-into-browsers direction :-)
  29. # [01:45] <Philip> (Currently "more time" means "still no time since I've got too many other things to do first", but anyway...)
  30. # [01:46] <mjs> Philip: I have to figure out how to get your test cases into the WebKit layout test framework
  31. # [01:47] <Philip> mjs: If there's something I can do to help with that, I'd be happy to attempt to find time to do so
  32. # [01:48] <Lachy> if anyone missed out on seeing the previous demo video, I've moved them to a new permanent location
  33. # [01:48] <Lachy> http://lachy.id.au/dev/markup/examples/video/
  34. # [01:48] <mjs> Philip: for our layout test format, any test that can show a difference in text content, layout metrics or style is best, so the red-green ones will integrate trivially
  35. # [01:48] <Philip> I really need to go through all the tests some time to make sure they're not broken and reflect the current/proposed spec and aren't missing certain cases
  36. # [01:49] <mjs> Philip: the blue ones that require inspection will only really work as pixel tests so won't be quite as great, but it would be hard to do fuller automation without having getImageData()
  37. # [01:49] <mjs> however, I must admit this is a motivation to add getImageData()
  38. # [01:50] <Philip> I guess I need to think about colour space issues some time too, because currently I'm assuming everything is easy and linear and matches the rendering of the PNGs, which might not be a safe assumption
  39. # [01:52] <Philip> Oh, and I need to think about how to test optional features (gradients, shadows), or propose that they be made non-optional
  40. # [01:53] <Philip> Can't think of much else, though
  41. # [01:55] <Philip> Oh, and I should write patches for browsers that are too lazy to fix the bugs themselves ;-)
  42. # [01:55] <gavin> Philip: it'd sure be great to get your test in Mozilla's automated test suite as well :)
  43. # [01:57] <Dashiva> mjs: But who will test getImageData()? :P
  44. # [01:57] <mjs> Dashiva: it's true that both drawing wrong and having a corresponding bug in getImageData() could make tests accidentally pass, but it seems pretty unlikelly
  45. # [01:58] <mjs> Philip: for WebKit, it's almost certain <canvas> rendering will remain in line with the way an <img> containing untagged image data renders
  46. # [01:58] <Philip> gavin: I'd be happy to help with that too, particularly if someone else did any boring / time-consuming bits :-)
  47. # [01:58] <mjs> Philip: also, patches very welcome, including patches to add some or all of the test suite to our regression tests
  48. # [02:00] <Philip> Dashiva: You can test getImageData by just drawing stuff and asserting the result values - the real problem is testing that the rendered output is not upside-down or byte-flipped compared to the internal bitmap, and that needs human inspection
  49. # [02:01] <Philip> My test code already has workarounds for two getImageData bugs (Firefox returning premultiplied alpha, Opera returning BGRA), so most of the automatically-checked tests work even when the getImageData tests fail
  50. # [02:01] <gavin> Philip: I'm not all that familiar with your tests, but it sounds like they could probably be ported to our mochitest-based framework fairly easily
  51. # [02:01] <gavin> or perhaps to the "reftest" framework
  52. # [02:02] <gavin> http://developer.mozilla.org/en/docs/Developing_Tests has some info, if you want to look into it :)
  53. # [02:06] * Quits: jgraham_ (james@81.86.218.70) (Quit: This computer has gone to sleep)
  54. # [02:08] * Quits: tH (Rob@87.102.21.203) (Quit: ChatZilla 0.9.79-rdmsoft [XULRunner 1.8.0.9/2006120508])
  55. # [02:11] <Philip> gavin: I expect most would work as MochiTests (though I've no idea how easy it would be to integrate with that), and only a few would have to be reftests instead
  56. # [02:12] <Philip> gavin: Some tests depend on accessing things across subdomains (to make sure security checks work sensibly) - is there any way to run those nicely, without depending on an actual external network connection?
  57. # [02:13] <gavin> yes, we have support for cross-domain testing
  58. # [02:13] * Quits: MikeSmith (MikeSmith@mcclure.w3.org) (Quit: Less talk, more pimp walk.)
  59. # [02:13] <gavin> using a custom HTTP server implemented in JS
  60. # [02:14] <Philip> Okay, sounds handy
  61. # [02:14] <gavin> http://developer.mozilla.org/en/docs/Mochitest#How_do_I_test_issues_which_only_show_up_when_tests_are_run_across_domains.3F
  62. # [02:15] <mjs> we have cross-domain tests using apache
  63. # [02:15] <mjs> (our regression test setup launches apache)
  64. # [02:15] <Philip> Hmm... If a page on http://example.com loads an image from ftp://example.com, does that count as being cross domain?
  65. # [02:16] <mjs> we really have only one kind of test
  66. # [02:16] <mjs> Philip: normally cross-scheme access is disallowed
  67. # [02:16] * Quits: myakura (myakura@122.29.186.5) (Quit: Leaving...)
  68. # [02:16] <mjs> so cross-domain isn't quite the most accurate term
  69. # [02:16] <mjs> browser security people usually say same-origin
  70. # [02:16] <mjs> (well, the good ones, the bad ones say cross-site scripting, which is confusing, since that often refers to server-side vulnerabilities too)
  71. # [02:17] <Philip> Aha, "origin" is the word I wasn't thinking of but should have been
  72. # [02:19] <Philip> I guess it might be nice to have a single source for all the canvas tests, and automatically construct specialised versions for each user (WebKit, Mozilla, etc), so updates can be more easily propagated to everyone
  73. # [02:20] <gavin> yeah, that would be ideal
  74. # [02:20] <gavin> dunno how hard it would be
  75. # [02:20] <gavin> I know we've imported some external testsuites with various rewriting hacks in the past
  76. # [02:20] <Philip> That makes it hard for the individual projects to make modifications, though, since they'd have to get the single source updated
  77. # [02:20] <gavin> but I think those were one time imports
  78. # [02:22] * Quits: Sander (svl@86.87.68.167) (Quit: And back he spurred like a madman, shrieking a curse to the sky.)
  79. # [02:23] * Philip generates the test files from a mildly crazy YAML + Python system, so it ought to be relatively straightforward to have various customised test case formats
  80. # [02:32] <Philip> gavin: The documentation makes it sound like Mochitest is quite a heavyweight mechanism - if I just want to run a bit of HTML and JS and output a boolean pass/fail, is it still the best thing to use?
  81. # [02:36] <gavin> writing a mochitest that does that isn't too heavyweight
  82. # [02:36] <gavin> let me find you an example
  83. # [02:38] <gavin> http://lxr.mozilla.org/seamonkey/source/testing/mochitest/tests/test_bug696.html
  84. # [02:38] <Philip> http://developer.mozilla.org/en/docs/Mochitest#Try_not_to_use_Mochitest sounds like it's trying to put me off :-)
  85. # [02:38] <gavin> there's some boilerplate overhead that kind of makes it look complicated
  86. # [02:39] <gavin> it is trying to put you off, if there's a better alternative :)
  87. # [02:39] <gavin> in this case I don't think there is
  88. # [02:39] <gavin> reftest could work, but then you need to provide reference renderings somehow, and for canvas that might be hard in some cases
  89. # [02:39] <Philip> Okay, thanks :-)
  90. # [02:41] <Philip> I have PNG reference images for most of the tests, but it's hard to get pixel-perfect matches since <canvas> isn't designed for that precision, so I expect it wouldn't work so well
  91. # [02:42] <gavin> yeah, could be
  92. # [02:42] <Philip> (and so lots of the tests use getImageData to extract a few pixels, and check the result is +/- 10% of the expected value, which stops most false negatives and doesn't let through too many false positives)
  93. # [02:43] * Quits: heycam (cam@203.217.79.225) (Ping timeout)
  94. # [02:49] * Joins: heycam (cam@203.217.91.18)
  95. # [03:17] * Joins: Zeros (Zeros-Elip@69.140.40.140)
  96. # [04:45] <mjs> getImageData is great for testability
  97. # [04:45] <mjs> kinda makes me want to implement it
  98. # [05:11] * Quits: mjs (mjs@64.81.48.145) (Quit: mjs)
  99. # [05:14] * Joins: mjs (mjs@64.81.48.145)
  100. # [05:24] * Joins: aroben (adamroben@67.160.250.192)
  101. # [06:38] * Quits: Zeros (Zeros-Elip@69.140.40.140) (Quit: Leaving)
  102. # [07:28] * Quits: sbuluf (rimz@200.49.132.89) (Ping timeout)
  103. # [07:50] * Joins: sbuluf (dmgfopl@200.49.132.97)
  104. # [07:51] * Quits: sbuluf (dmgfopl@200.49.132.97) (Quit: sbuluf)
  105. # [07:53] * Joins: sbuluf (hwdp@200.49.132.87)
  106. # [08:38] * Joins: shepazu_ (schepers@128.30.52.30)
  107. # [08:38] * Quits: shepazu (schepers@128.30.52.30) (Ping timeout)
  108. # [09:40] * Quits: sbuluf (hwdp@200.49.132.87) (Ping timeout)
  109. # [09:45] * Parts: anne (annevk@82.156.27.18)
  110. # [09:48] * Quits: aroben (adamroben@67.160.250.192) (Connection reset by peer)
  111. # [09:48] * Joins: aroben (adamroben@67.160.250.192)
  112. # [10:10] * Joins: ROBOd (robod@89.122.216.38)
  113. # [10:15] * Joins: jgraham_ (james@81.86.218.70)
  114. # [10:48] * Quits: aroben (adamroben@67.160.250.192) (Quit: aroben)
  115. # [12:00] * Joins: gsnedders (gsnedders@86.145.188.131)
  116. # [12:21] * Joins: sbuluf (pzamdx@200.49.132.103)
  117. # [12:21] * Quits: sbuluf (pzamdx@200.49.132.103) (Quit: sbuluf)
  118. # [12:21] * Joins: sbuluf (zblzp@200.49.132.103)
  119. # [14:07] * Joins: tH_ (Rob@87.102.21.203)
  120. # [14:07] * tH_ is now known as tH
  121. # [14:30] * Joins: Sander (svl@86.87.68.167)
  122. # [14:44] * Quits: Lachy (Lachlan@84.215.41.149) (Quit: Leaving)
  123. # [14:45] * Joins: Lachy (Lachlan@84.215.41.149)
  124. # [15:07] <Philip> mjs: I certainly won't complain if you want to implement getImageData :-)
  125. # [15:08] <Philip> Multiple bonus points go to anyone who can implement it for Excanvas in IE
  126. # [15:12] * Quits: Lachy (Lachlan@84.215.41.149) (Quit: This computer has gone to sleep)
  127. # [15:13] <Philip> anne: I've added some notes to http://wiki.whatwg.org/wiki/SVG_and_canvas
  128. # [15:51] * Joins: Lachy (Lachlan@88.91.106.25)
  129. # [16:05] * Quits: Sander (svl@86.87.68.167) (Quit: And back he spurred like a madman, shrieking a curse to the sky.)
  130. # [16:21] * Joins: anne-mac (annevk@80.202.68.140)
  131. # [16:23] * Joins: myakura (myakura@122.29.186.5)
  132. # [16:26] <anne-mac> thanks Philip, that's useful :)
  133. # [16:26] * Philip notices that http://www.w3.org/TR/SVG-access/ uses SVG in no namespace (except in the final example)
  134. # [16:26] <Philip> (Is that just an error, or is it correct according to an old/obsolete version of SVG?)
  135. # [16:27] <Philip> anne-mac: I expect there are more things to say but I can't think of any more right now :-)
  136. # [16:29] <anne-mac> old obsolete version of SVG says Charles
  137. # [16:29] <anne-mac> he notes that the DTD reference also says "CR"
  138. # [16:33] <Philip> Ah, http://www.w3.org/TR/2000/CR-SVG-20000802/struct.html#DOMInterfaces seems to indicate you have to use <svg:svg> when embedding in another document, but can use plain <svg> for the root element of an SVG document
  139. # [16:48] * Quits: anne-mac (annevk@80.202.68.140) (Ping timeout)
  140. # [17:00] * Quits: gsnedders (gsnedders@86.145.188.131) (Quit: gsnedders)
  141. # [17:06] * Quits: myakura (myakura@122.29.186.5) (Quit: Leaving...)
  142. # [17:11] * Joins: Sander (svl@86.87.68.167)
  143. # [17:19] * Joins: gsnedders (gsnedders@86.145.188.131)
  144. # [18:18] * Quits: ChrisWilson (cwilso@131.107.0.104) (Ping timeout)
  145. # [18:23] * Joins: ChrisWilson (cwilso@131.107.0.103)
  146. # [18:40] * Joins: aroben (adamroben@67.160.250.192)
  147. # [18:50] * Joins: anneruneh (opera@80.203.30.152)
  148. # [19:11] * Quits: aroben (adamroben@67.160.250.192) (Ping timeout)
  149. # [19:22] * Quits: Lachy (Lachlan@88.91.106.25) (Quit: This computer has gone to sleep)
  150. # [19:27] * Joins: zcorpan (zcorpan@83.227.33.203)
  151. # [19:47] * Quits: zcorpan (zcorpan@83.227.33.203) (Ping timeout)
  152. # [19:52] * Joins: Lachy (Lachlan@80.203.102.149)
  153. # [19:56] * Quits: Lachy (Lachlan@80.203.102.149) (Ping timeout)
  154. # [19:57] * Quits: anneruneh (opera@80.203.30.152) (Ping timeout)
  155. # [20:13] * Joins: aroben (adamroben@67.160.250.192)
  156. # [21:11] * Joins: Lachy (Lachlan@80.203.30.152)
  157. # [21:16] * Quits: aroben (adamroben@67.160.250.192) (Quit: aroben)
  158. # [21:25] * Joins: hober (ted@68.107.112.172)
  159. # [21:30] * Quits: ROBOd (robod@89.122.216.38) (Quit: http://www.robodesign.ro )
  160. # [21:34] * Quits: Lachy (Lachlan@80.203.30.152) (Quit: Leaving)
  161. # [22:16] * Joins: zcorpan (zcorpan@83.227.33.203)
  162. # [22:48] * Quits: zcorpan (zcorpan@83.227.33.203) (Ping timeout)
  163. # [23:01] <Philip> Mochitest says "Passed: 4. Failed: 0", which sounds good so far
  164. # [23:40] * Joins: dbaron (dbaron@71.204.145.103)
  165. # [23:51] * Joins: heycam` (cam@130.194.72.84)
  166. # Session Close: Mon Dec 03 00:00:00 2007

The end :)