97281 2012-09-20 19:03:20 -0700 [Chromium] Use harfbuzz-ng by default on Linux 2012-11-05 11:21:45 -0800 1 1 1 Unclassified WebKit Platform 528+ (Nightly build) Unspecified Unspecified RESOLVED FIXED P2 Normal --- 97993 72984 1 bashi bashi behdad dglazkov evan jshin tony webkit.review.bot oldest_to_newest 725074 0 bashi 2012-09-20 19:03:20 -0700 As of r129175, I think harfbuzz-ng and HarfBuzzShaper are ready to use by default. - Layout tests results are slightly differ from old harfbuzz, but most of them are improved. - Page cycler intl2 results are bit regressed, but we should not value the result too much. Detailed explanations will follow. 725076 1 bashi 2012-09-20 19:05:19 -0700 Here is the LayoutTests results: https://docs.google.com/open?id=0B6NYyLPujP4TNmJjS0JmbWVKRlU Summary of results (only complex text related tests): - Looks improved. fast/text/cg-vs-atsui.html fast/text/complex-text-opacity.html fast/text/atsui-negative-spacing-features.html fast/text/atsui-spacing-features.html fast/text/international/hindi-spacing.html fast/text/international/text-spliced-font.html fast/text/international/thai-line-breaks.html platform/chromium-linux/fast/text/international/arabic-vertical-offset.html platform/chromium-linux/fast/text/international/complex-joining-using-gpos.html - Looks fine. fast/text/drawBidiText.html fast/text/international/arabic-justify.html fast/text/international/hebrew-vowels.html Tests under svg/W3C-I18N/ svg/W3C-SVG-1.1-SE/text-intro-05-t.svg svg/W3C-SVG-1.1/text-intro-05-t.svg svg/text/bidi-text-query.svg svg/text/text-intro-05-t.svg transforms/2d/hindi-rotated.html - Just need rebaselines. fast/text/international/khmer-selection.html platform/chromium-linux/svg/text/text-with-geometric-precision.svg 725081 2 bashi 2012-09-20 19:07:09 -0700 Page cycler intl2 summary (20 iterations): - hb-old milliseconds: 61405 mean per set: 3070.25 mean per page: 102.34 timer lag: 11499.00 timer lag per page: 19.17 - hb-ng milliseconds: 64953 mean per set: 3247.65 mean per page: 108.26 timer lag: 11933.00 timer lag per page: 19.89 hb-old a bit fast, but we shouldn't value it too much. Below is a quote from an email from behdad@ about the result: <quote> One thing concerns me regarding both these numbers, and LayoutTests. Looks like all these tests are running against the WinXP set of fonts which are extremely incomplete in their OpenType features. In particular, Times New Roman from XP doesn't have GDEF or GPOS. So, we end up hitting a lot of fallback code that is rarely exercised these days. So, IMO, the numbers are a bit misleading and irrelevant. With modern fonts I expect new HB to perform better than old HB. </quote> See also http://crbug.com/139920. 725084 3 bashi 2012-09-20 19:11:30 -0700 Hi Tony, I'd like ask your opinion about using harfbuzz-ng by default on Linux. I put the current status and summary. What do you think? 725638 4 evan 2012-09-21 11:01:48 -0700 Tony asked me what I think. 1) It'd be nice to eliminate all this old code, so I'm tempted to say we should do it, BUT 2) Why do we even have performance tests if we're not going to believe the results? Behdad's comment means that either: 2a) The tests are using fonts that don't match what normal users have; or 2b) Normal users are going to get 5% slower. It would be nice to understand which of those is actually true. I think 2b is probably fine but we should flip the flag knowing that we're losing performance if that's what it takes. 725641 5 tony 2012-09-21 11:08:35 -0700 Can you re-run the perf tests using a different font for both harfbuzz-old and harfbuzz-ng? For example, you could replace Times New Roman as the default font with "serif" to get the fontconfig default (looks like DejaVu on my system). You could also try using the ChromeOS fonts (Droid, I think). We can find out if our users have Times New Roman installed or not using UMA. It would be nice to know how many users are using Times New Roman vs something else. 725643 6 evan 2012-09-21 11:11:14 -0700 To clarify: I'm not even in a position of influence on this project anymore, so this is just my opinion. And my opinion is that it's totally fine to lose 5% of performance if it means deleting a bunch of crufty code that doesn't quite work right. 725920 7 bashi 2012-09-21 16:02:40 -0700 Hi Evan, Tony, Thank you so much for your inputs. I'll re-run page cycler tests using different font next Monday. I'll also check UMA to have ideas to answer Evan's question. (I'm not sure how I can see UMA, but I guess I can ask someone to see it) 725995 8 jshin 2012-09-21 17:18:22 -0700 > 2a) The tests are using fonts that don't match what normal users have I responded to the quoted comment of Behdad in the email thread among him, bashi and me, but bashi only quoted Behdad's first comment. The answer is not that simple. It all depends on what portion of our users have Arial and Times New Roman (a very old version dating from 2000) as included in mscorettf package. If the majority of our users have those old Arial and Times New Roman, our perf bot test environment is realistic. Even if that's the case, Arial and Times New Roman as included in mscorettf covers only Arabic and Hebrew among complex scripts tested by intl2 page_cycler tests. So, Thai and Indic script pages should not be affected by what Behdad mentioned. It means that the overall perf regression should come from Arabic/Hebrew portion alone and it's actually more worrisome because the perf regression will be more noticeable for Arabic/Hebrew pages. If that's the case, we do now have a better alternative (Arimo, Tinos) to Arial/Times New Roman (dating from 2000) and we can evangelize using those fonts in place of Arial/Times New Roman. If the majority of users do not have Arial / Times New Roman (from 2000), Arabic and Hebrew will be rendered with modern fonts and hb-ng's fallback code (which is triggered only when appropriate GSUB/GPOS tables are absent) will not be used. In that case, our buildbot does not reflect the reality very well and we don't have to worry about this. > Can you re-run the perf tests using a different font for both harfbuzz-old and harfbuzz-ng? > For example, you could replace Times New Roman as the default font with "serif" to get the fontconfig > default (looks like DejaVu on my system). You could also try using the ChromeOS fonts (Droid, I think). ChromeOS fonts corresponding to Arial and Times New Roman are Arimo and Tinos. Just adding a fontconfig substitution directive alone would not work because we require the exact name match. Fortunately for Arimo and Tinos, we already hardcode them to be equivalent to Arial and Times New Roman. So, what's necessary is to uninstall mscorettf package and to add a fontconfig line to make them (Arimo, Tinos) equivalent to Arial and Tinos. local.conf file in fontconfig ebuild for ChromeOS has entries necessary for that. 725997 9 jshin 2012-09-21 17:21:11 -0700 > And my opinion is that it's totally fine to lose 5% of performance if it means deleting a bunch of > crufty code that doesn't quite work right. I also agree to the above despite what I wrote about Arabic/Hebrew. The correctness is more important than ~5% perf gain for an unknown percentage of users (who have 12-yr old version of Arial and Times New roman). 726330 10 bashi 2012-09-23 18:51:44 -0700 Jungshik, I'm sorry for not quoting your reply. I should mentioned it. And thank you for the detailed follow-ups. I ran page cycler intl2 tests using alternative fonts(Arimo and Tinos): - hb-old milliseconds 67738 mean per set 3386.90 mean per page 112.90 timer lag 13158.00 timer lag per page 21.93 - hb-ng milliseconds 72671 mean per set 3633.55 mean per page 121.12 timer lag 13308.00 timer lag per page 22.18 Looks like ToT is somewhat degraded (not sure, there might be another reason on my machine). Anyway, it seems that the performance loss is irrelevant to which fonts are used. We should improve the performance, but I'd like attach importance to correctness at this time. I and behdad will address the performance as an ongoing work. If it makes sense, I'll create a patch to make hb-ng by default. 726841 11 tony 2012-09-24 10:57:02 -0700 A 7% regression seems like a lot. I bet there's some low hanging fruit. Can we profile/optimize first then revisit the question of enabling by default? I think if we switch now, there will be no incentive to make things faster. 726978 12 behdad 2012-09-24 13:59:20 -0700 (In reply to comment #11) > A 7% regression seems like a lot. I bet there's some low hanging fruit. Can we profile/optimize first then revisit the question of enabling by default? Ok, that is indeed cause for concern. In testing HarfBuzz alone I have not been able to reproduce this. So, my gut feeling is that it's coming from the webkit integration layer. That said, here's how we can move forward: 1) Confirm whether my results are valid. Ie. whether we can reproduce the slowdown in HarfBuzz along (with old vs new backends) on someone else's testing (bashi?). 2) Updating to ICU 50 should give harfbuzz-ng some speedup. Jungshik, any update on that? Alternatively, since we are not relying on the normalization features of hb-ng in Chrome yet, maybe we can disable the normalizer for some speedup. 3) Good old profiling. We may be doing something differently in the callbacks for example. It's hard to know what's going on right now. > I think if we switch now, there will be no incentive to make things faster. While that's generally true, I don't think that would be the case here. We're removing code that has been unmaintained for many years, and replacing with something that my fulltime job to work on, and is also used in GNOME and Firefox. It's a major part of my job for the next two years to improve ChromeOS's text rendering, and I know bashi also has similar plans. BTW, there's also memory savings to be had by the switch. Checkout: http://goo.gl/woyty Someone should confirm that we observe this in Chrome. 726995 13 behdad 2012-09-24 14:19:39 -0700 As another example of why switching over sooner than later is a good idea: Myanmar seems to be broken with old HarfBuzz but works fine with new. This website for example: http://mm.myanmars.net/ (Uses a webfont.) 727526 14 bashi 2012-09-25 05:12:49 -0700 (In reply to comment #11) > A 7% regression seems like a lot. I bet there's some low hanging fruit. Can we profile/optimize first then revisit the question of enabling by default? > > I think if we switch now, there will be no incentive to make things faster. I'm trying to profile with --single-process as http://www.chromium.org/developers/profiling-chromium-and-webkit suggests. It seems that fonts for some scripts (Thai, Devenagari, etc) aren't available with --single-process so I'm looking for a better way to profile complex text path rendering. 728980 15 behdad 2012-09-26 14:31:31 -0700 (In reply to comment #14) > (In reply to comment #11) > > A 7% regression seems like a lot. I bet there's some low hanging fruit. Can we profile/optimize first then revisit the question of enabling by default? > > > > I think if we switch now, there will be no incentive to make things faster. > > I'm trying to profile with --single-process as http://www.chromium.org/developers/profiling-chromium-and-webkit suggests. It seems that fonts for some scripts (Thai, Devenagari, etc) aren't available with --single-process so I'm looking for a better way to profile complex text path rendering. Sounds weird. Did you make any progress here? 728981 16 behdad 2012-09-26 14:31:58 -0700 Tony, Evan, given the discussion, can we get an updated opinion from you two? 728987 17 tony 2012-09-26 14:36:53 -0700 My opinion is still the same. I would like to see some profiling/optimization before switching. We don't have to get back to the same speed, you just need to show me that you're interested in making things faster (so far, it seems very hand-wavy). 728991 18 evan 2012-09-26 14:39:28 -0700 Unfortunately --single-process turns off a lot of code, including some of the code related to complex text. (As I recall it's because we have to run a separate subprocess to make fontconfig calls as it has global data structures and the Chrome UI thread also can make fontconfig calls.) 729003 19 behdad 2012-09-26 14:56:15 -0700 bashi, since getting profiling seems to be challenging, can you make sure we're not doing excessive work? Ie. count the number of face and font creations as well as shape calls for old / new harfbuzz? I'm suspecting that we're doing something wrong. 729006 20 behdad 2012-09-26 15:01:43 -0700 FWIW, I did some fresh measurements with hb-shape with WinXP and Win7 Times New Roman and long Arabic text, and I see solid improvements in new HB. Here's my test command: $ time ./hb-shape ~/fonts/WinXP/times.ttf --text-file test-file --output-file /dev/null --no-glyph-names --shaper {ot|old} (choose one shaper) I performed two tests, using text files having: * 10,000 lines of the same 24k-byte Persian text, * 1000,000 lines of the same 10-byte Persian text, In both cases, for WinXP times.ttf I see new HB faster than old by about 20%, and with Win7 times.ttf I see it 35% faster. Given the Arabic numbers we are observing in page cycler, I'm sure we're doing something wrong in the webkit integration patch. 729010 21 tony 2012-09-26 15:03:41 -0700 (In reply to comment #20) > > In both cases, for WinXP times.ttf I see new HB faster than old by about 20%, and with Win7 times.ttf I see it 35% faster. Given the Arabic numbers we are observing in page cycler, I'm sure we're doing something wrong in the webkit integration patch. Great. We should fix this before turning on harfbuzz-ng. 729023 22 bashi 2012-09-26 15:37:26 -0700 HarfBuzzShaper(hb-ng) do more work than ComplexTextController(hb-old) for correct shaping (e.g. positioning and dividing text run). I doubt we are doing something wrong. At least I'm sure we don't create extra hb_font and hb_face objects. Anyway I agree with we need to fix performance issue. I'm swampwed other tasks now, but I will try make time for this. 729068 23 jshin 2012-09-26 16:42:30 -0700 (In reply to comment #12) > 2) Updating to ICU 50 should give harfbuzz-ng some speedup. Jungshik, any update on that? ICU 50 hasn't yet been released :-) It's supposed to be released on Oct 31st. > Alternatively, since we are not relying on the normalization features of hb-ng in Chrome yet, maybe we can disable the normalizer for some speedup. Have we tried that to see if it helps? 729091 24 bashi 2012-09-26 17:20:05 -0700 (In reply to comment #23) > > Alternatively, since we are not relying on the normalization features of hb-ng in Chrome yet, maybe we can disable the normalizer for some speedup. > > Have we tried that to see if it helps? I haven't. I'll try. BTW, normalization on HarfBuzzShaper and ComplexTextController is a bit different. hb-old: http://code.google.com/searchframe#OAMlx_jo-ck/src/third_party/WebKit/Source/WebCore/platform/graphics/harfbuzz/HarfBuzzShaperBase.cpp&exact_package=chromium&type=cs&l=55 hb-ng: http://code.google.com/searchframe#OAMlx_jo-ck/src/third_party/WebKit/Source/WebCore/platform/graphics/harfbuzz/ng/HarfBuzzShaper.cpp&type=cs&l=162 729150 25 bashi 2012-09-26 19:28:39 -0700 (In reply to comment #24) > > Have we tried that to see if it helps? Not calling normalizedCharacters() gains 2%, but '\t', '\n', and U+00a0 are displayed with square boxes. These characters should be replaced with spaces in WebKit as normalizedCharacters() did. We can't remove this conversion and get the gain. 729826 26 behdad 2012-09-27 13:45:01 -0700 Just to rule out some possibility: what do you get if you use hb_unicode_funcs_get_empty() instead of hb_icu_get_unicode_funcs()? Results should most be wrong, but I want to see the perf characteristics. 738519 27 bashi 2012-10-09 21:44:12 -0700 The current performance of page cycler intl2 tests (measured by using WebKit r130840). - hb-old iterations 5 pages 30 milliseconds 19620 mean per set 3924.00 mean per page 130.80 timer lag 3278.00 timer lag per page 21.85 iterations 20 pages 30 milliseconds 68783 mean per set 3439.15 mean per page 114.64 timer lag 12589.00 timer lag per page 20.98 - hb-ng iterations 5 pages 30 milliseconds 20097 mean per set 4019.40 mean per page 133.98 timer lag 3134.00 timer lag per page 20.89 iterations 20 pages 30 milliseconds 70054 mean per set 3502.70 mean per page 116.76 timer lag 11702.00 timer lag per page 19.50 hb-ng is slightly slow. Tony, do you think we should make further improvement? 738545 28 bashi 2012-10-09 22:08:43 -0700 (In reply to comment #26) > Just to rule out some possibility: what do you get if you use hb_unicode_funcs_get_empty() instead of hb_icu_get_unicode_funcs()? Results should most be wrong, but I want to see the perf characteristics. Here is the result. iterations 5 pages 30 milliseconds 19376 mean per set 3875.20 mean per page 129.17 timer lag 3083.00 timer lag per page 20.55 iterations 20 pages 30 milliseconds 65751 mean per set 3287.55 mean per page 109.59 timer lag 12176.00 timer lag per page 20.29 It actually became faster, but it also implied that the bottlenecks are in my WebKit side.. BTW, here are some notes about performance regression. - hb-old implementation does NFC normalization in advance. This contributes to performance. But it causes wrong selection behavior because the character index will be mismatched between the original TextRun and normalized character sequence. For example, you can see an wrong behavior on fast/text/atsui-kerning-and-ligatures.html (Please try to select amèt from left to right) hb-ng implementation fixed this issue, but we needed to add further checks whether the font to be used actually has the glyphs for combining characters. - hb-old implementation directly calls Skia APIs. On the other hand, hb-ng implementation fills glyphs and positions into GlyphBuffer and calls port-independent function to draw glyphs. This is a design choice to make hb-ng implementation be Skia-independent. However, this might affect the performance. 739057 29 tony 2012-10-10 10:12:16 -0700 This looks much better! We're down to a 2-2.5% regression. Yes, let's switch. Thank you for profiling and finding some low hanging fruit first. 740360 30 168301 bashi 2012-10-11 16:04:33 -0700 Created attachment 168301 Patch 740367 31 bashi 2012-10-11 16:08:26 -0700 Hi Tony, Thank you so much for helping this transition. I'm preparing a patch to switch. Should I also update test expectations in this patch? Or just adding suppressions in TestExpectations and then updating expectations by using garden-o-matic is ok? 740404 32 168313 bashi 2012-10-11 16:35:23 -0700 Created attachment 168313 Add temporary expectations 740439 33 168313 tony 2012-10-11 16:54:55 -0700 Comment on attachment 168313 Add temporary expectations Suppressing in TestExpectations and rebaselining using garden-o-matic is fine. Please watch the perf bots and be ready to update the expectations file if they go red. 740454 34 bashi 2012-10-11 17:12:59 -0700 (In reply to comment #33) > (From update of attachment 168313 [details]) > Suppressing in TestExpectations and rebaselining using garden-o-matic is fine. Please watch the perf bots and be ready to update the expectations file if they go red. I see. I'll keep watch on perf bots after landing. I'll cq+ later so that I can work together with non-PST gardener. I guess perf bots may go red. If the regression rate is within a expected range(~2%), I'll ask perf sheriffs whether we can update the perf expectation. 740480 35 168313 webkit.review.bot 2012-10-11 17:55:23 -0700 Comment on attachment 168313 Add temporary expectations Attachment 168313 did not pass chromium-ews (chromium-xvfb): Output: http://queues.webkit.org/results/14262313 740570 36 168313 webkit.review.bot 2012-10-11 20:36:30 -0700 Comment on attachment 168313 Add temporary expectations Clearing flags on attachment: 168313 Committed r131134: <http://trac.webkit.org/changeset/131134> 740571 37 webkit.review.bot 2012-10-11 20:36:34 -0700 All reviewed patches have been landed. Closing bug. 168301 2012-10-11 16:04:33 -0700 2012-10-11 16:35:18 -0700 Patch bug-97281-20121012080330.patch text/plain 5265 bashi U3VidmVyc2lvbiBSZXZpc2lvbjogMTMxMDk4CmRpZmYgLS1naXQgYS9Tb3VyY2UvV2ViQ29yZS9D aGFuZ2VMb2cgYi9Tb3VyY2UvV2ViQ29yZS9DaGFuZ2VMb2cKaW5kZXggOTJmM2EyZDA5MTUyMTBl ZTRiMjY3MGUyOTNlZmRkY2NkMWM4NzZlYy4uMzM4YjRkYmE1NGQzMWU5NjRiYzJlY2MwNjNjYWIx YzIzMDU3NTk5ZCAxMDA2NDQKLS0tIGEvU291cmNlL1dlYkNvcmUvQ2hhbmdlTG9nCisrKyBiL1Nv dXJjZS9XZWJDb3JlL0NoYW5nZUxvZwpAQCAtMSwzICsxLDE3IEBACisyMDEyLTEwLTExICBLZW5p Y2hpIElzaGliYXNoaSAgPGJhc2hpQGNocm9taXVtLm9yZz4KKworICAgICAgICBbQ2hyb21pdW1d IFVzZSBoYXJmYnV6ei1uZyBieSBkZWZhdWx0IG9uIExpbnV4CisgICAgICAgIGh0dHBzOi8vYnVn cy53ZWJraXQub3JnL3Nob3dfYnVnLmNnaT9pZD05NzI4MQorCisgICAgICAgIFJldmlld2VkIGJ5 IE5PQk9EWSAoT09QUyEpLgorCisgICAgICAgIFN3aXRjaCBvbGQgaGFyZmJ1enogdG8gaGFyZmJ1 enotbmcgb24gY2hyb21pdW0gbGludXggcG9ydC4KKworICAgICAgICBObyBuZXcgdGVzdHMuIEV4 aXN0aW5nIHRlc3RzIHNob3VsZCBjb3ZlciB0aGlzIHN3aXRjaC4KKyAgICAgICAgVGVzdCBleHBl Y3RhdGlvbnMgd2lsbCBiZSB1cGRhdGVkLgorCisgICAgICAgICogV2ViQ29yZS5neXAvV2ViQ29y ZS5neXA6CisKIDIwMTItMTAtMTEgIEpvc2h1YSBCZWxsICA8anNiZWxsQGNocm9taXVtLm9yZz4K IAogICAgICAgICBJbmRleGVkREI6IElETCB0eXBlcyBkZWZpbmVkIGluIHNwZWMgc2hvdWxkIGJl IHZpc2libGUgdG8gc2NyaXB0cwpkaWZmIC0tZ2l0IGEvU291cmNlL1dlYktpdC9jaHJvbWl1bS9D aGFuZ2VMb2cgYi9Tb3VyY2UvV2ViS2l0L2Nocm9taXVtL0NoYW5nZUxvZwppbmRleCA4MDNmMDIz YjcwOWY4NWFlMzA1MTk0MGFlNGZjYjRjY2UzYWU1NWUxLi5jZGI4N2YzYWY4NWY4ODA3ZGEyOGVh M2ExMjg1YWE2ZWI0ODEyMzhmIDEwMDY0NAotLS0gYS9Tb3VyY2UvV2ViS2l0L2Nocm9taXVtL0No YW5nZUxvZworKysgYi9Tb3VyY2UvV2ViS2l0L2Nocm9taXVtL0NoYW5nZUxvZwpAQCAtMSwzICsx LDEyIEBACisyMDEyLTEwLTExICBLZW5pY2hpIElzaGliYXNoaSAgPGJhc2hpQGNocm9taXVtLm9y Zz4KKworICAgICAgICBbQ2hyb21pdW1dIFVzZSBoYXJmYnV6ei1uZyBieSBkZWZhdWx0IG9uIExp bnV4CisgICAgICAgIGh0dHBzOi8vYnVncy53ZWJraXQub3JnL3Nob3dfYnVnLmNnaT9pZD05NzI4 MQorCisgICAgICAgIFJldmlld2VkIGJ5IE5PQk9EWSAoT09QUyEpLgorCisgICAgICAgICogZmVh dHVyZXMuZ3lwaTogRW5hYmxlIFdURl9VU0VfSEFSRkJVWlpfTkcgd2hlbiB1c2VfeDExID09IDEu CisKIDIwMTItMTAtMTEgIEFkYW0gQmFydGggIDxhYmFydGhAd2Via2l0Lm9yZz4KIAogICAgICAg ICBbQ2hyb21pdW1dIEltcHJvdmUgdGhlIGNvbW1lbnQgZGVzY3JpYmluZyBtYWluV29ybGRTY3Jp cHRDb250ZXh0CmRpZmYgLS1naXQgYS9Tb3VyY2UvV2ViQ29yZS9XZWJDb3JlLmd5cC9XZWJDb3Jl Lmd5cCBiL1NvdXJjZS9XZWJDb3JlL1dlYkNvcmUuZ3lwL1dlYkNvcmUuZ3lwCmluZGV4IGFmYmIz MWVjM2E3YjhlY2JmNzNkNDU0NWYzYzRkZTQxODQ1ZGFkMWYuLmMyYTgwNzg5NTJmNzcwMDEyODEw ODU3YTZkM2MzYTIwNmRlYTU2NDEgMTAwNjQ0Ci0tLSBhL1NvdXJjZS9XZWJDb3JlL1dlYkNvcmUu Z3lwL1dlYkNvcmUuZ3lwCisrKyBiL1NvdXJjZS9XZWJDb3JlL1dlYkNvcmUuZ3lwL1dlYkNvcmUu Z3lwCkBAIC00OCw4ICs0OCw2IEBACiAKICAgICAnZW5hYmxlX3dleGl0X3RpbWVfZGVzdHJ1Y3Rv cnMnOiAxLAogCi0gICAgJ3VzZV9oYXJmYnV6el9uZyUnOiAwLAotCiAgICAgJ3dlYmNvcmVfaW5j bHVkZV9kaXJzJzogWwogICAgICAgJy4uLycsCiAgICAgICAnLi4vLi4nLApAQCAtMjQ4LDcgKzI0 Niw3IEBACiAgICAgICAgICAgJy4uL3BsYXRmb3JtL2dyYXBoaWNzL2hhcmZidXp6JywKICAgICAg ICAgXSwKICAgICAgIH1dLAotICAgICAgWyd1c2VfeDExPT0xIGFuZCB1c2VfaGFyZmJ1enpfbmc9 PTEnLCB7CisgICAgICBbJ3VzZV94MTE9PTEnLCB7CiAgICAgICAgICd3ZWJjb3JlX2luY2x1ZGVf ZGlycyc6IFsKICAgICAgICAgICAnLi4vcGxhdGZvcm0vZ3JhcGhpY3MvaGFyZmJ1enovbmcnLAog ICAgICAgICBdLApAQCAtMTY2MywxMSArMTY2MSwxMiBAQAogICAgICAgICBbJ3VzZV94MTEgPT0g MScsIHsKICAgICAgICAgICAnc291cmNlcy8nOiBbCiAgICAgICAgICAgICAjIENoZXJyeS1waWNr IGZpbGVzIGV4Y2x1ZGVkIGJ5IHRoZSBicm9hZGVyIHJlZ3VsYXIgZXhwcmVzc2lvbnMgYWJvdmUu Ci0gICAgICAgICAgICBbJ2luY2x1ZGUnLCAncGxhdGZvcm0vZ3JhcGhpY3MvaGFyZmJ1enovQ29t cGxleFRleHRDb250cm9sbGVySGFyZkJ1enpcXC5jcHAkJ10sCiAgICAgICAgICAgICBbJ2luY2x1 ZGUnLCAncGxhdGZvcm0vZ3JhcGhpY3MvaGFyZmJ1enovRm9udEhhcmZCdXp6XFwuY3BwJCddLAog ICAgICAgICAgICAgWydpbmNsdWRlJywgJ3BsYXRmb3JtL2dyYXBoaWNzL2hhcmZidXp6L0ZvbnRQ bGF0Zm9ybURhdGFIYXJmQnV6elxcLmNwcCQnXSwKLSAgICAgICAgICAgIFsnaW5jbHVkZScsICdw bGF0Zm9ybS9ncmFwaGljcy9oYXJmYnV6ei9IYXJmQnV6elNraWFcXC5jcHAkJ10sCiAgICAgICAg ICAgICBbJ2luY2x1ZGUnLCAncGxhdGZvcm0vZ3JhcGhpY3MvaGFyZmJ1enovSGFyZkJ1enpTaGFw ZXJCYXNlXFwuKGNwcHxoKSQnXSwKKyAgICAgICAgICAgIFsnaW5jbHVkZScsICdwbGF0Zm9ybS9n cmFwaGljcy9oYXJmYnV6ei9uZy9IYXJmQnV6ek5HRmFjZVxcLihjcHB8aCkkJ10sCisgICAgICAg ICAgICBbJ2luY2x1ZGUnLCAncGxhdGZvcm0vZ3JhcGhpY3MvaGFyZmJ1enovbmcvSGFyZkJ1enpO R0ZhY2VTa2lhXFwuY3BwJCddLAorICAgICAgICAgICAgWydpbmNsdWRlJywgJ3BsYXRmb3JtL2dy YXBoaWNzL2hhcmZidXp6L25nL0hhcmZCdXp6U2hhcGVyXFwuKGNwcHxoKSQnXSwKICAgICAgICAg ICAgIFsnaW5jbHVkZScsICdwbGF0Zm9ybS9ncmFwaGljcy9vcGVudHlwZS9PcGVuVHlwZVR5cGVz XFwuaCQnXSwKICAgICAgICAgICAgIFsnaW5jbHVkZScsICdwbGF0Zm9ybS9ncmFwaGljcy9vcGVu dHlwZS9PcGVuVHlwZVZlcnRpY2FsRGF0YVxcLihjcHB8aCkkJ10sCiAgICAgICAgICAgICBbJ2lu Y2x1ZGUnLCAncGxhdGZvcm0vZ3JhcGhpY3Mvc2tpYS9TaW1wbGVGb250RGF0YVNraWFcXC5jcHAk J10sCkBAIC0xNjc4LDE1ICsxNjc3LDYgQEAKICAgICAgICAgICAgIFsnZXhjbHVkZScsICdIYXJm YnV6elteL10rXFwuKGNwcHxoKSQnXSwKICAgICAgICAgICBdLAogICAgICAgICB9XSwKLSAgICAg ICAgWyd1c2VfeDExPT0xIGFuZCB1c2VfaGFyZmJ1enpfbmc9PTEnLCB7Ci0gICAgICAgICAgJ3Nv dXJjZXMvJzogWwotICAgICAgICAgICAgWydleGNsdWRlJywgJ3BsYXRmb3JtL2dyYXBoaWNzL2hh cmZidXp6L0NvbXBsZXhUZXh0Q29udHJvbGxlckhhcmZCdXp6XFwuY3BwJCddLAotICAgICAgICAg ICAgWydleGNsdWRlJywgJ3BsYXRmb3JtL2dyYXBoaWNzL2hhcmZidXp6L0hhcmZCdXp6U2tpYVxc LmNwcCQnXSwKLSAgICAgICAgICAgIFsnaW5jbHVkZScsICdwbGF0Zm9ybS9ncmFwaGljcy9oYXJm YnV6ei9uZy9IYXJmQnV6ek5HRmFjZVxcLihjcHB8aCkkJ10sCi0gICAgICAgICAgICBbJ2luY2x1 ZGUnLCAncGxhdGZvcm0vZ3JhcGhpY3MvaGFyZmJ1enovbmcvSGFyZkJ1enpOR0ZhY2VTa2lhXFwu Y3BwJCddLAotICAgICAgICAgICAgWydpbmNsdWRlJywgJ3BsYXRmb3JtL2dyYXBoaWNzL2hhcmZi dXp6L25nL0hhcmZCdXp6U2hhcGVyXFwuKGNwcHxoKSQnXSwKLSAgICAgICAgICBdLAotICAgICAg ICB9XSwKICAgICAgICAgWyd0b29sa2l0X3VzZXNfZ3RrID09IDEnLCB7CiAgICAgICAgICAgJ3Nv dXJjZXMvJzogWwogICAgICAgICAgICAgIyBDaGVycnktcGljayBmaWxlcyBleGNsdWRlZCBieSB0 aGUgYnJvYWRlciByZWd1bGFyIGV4cHJlc3Npb25zIGFib3ZlLgpAQCAtMTY5Nyw3ICsxNjg3LDEy IEBACiAgICAgICAgICAgICBbJ2V4Y2x1ZGUnLCAnR3RrXFwuY3BwJCddLAogICAgICAgICAgIF0s CiAgICAgICAgIH1dLAotICAgICAgICBbJ3VzZV94MTE9PTEgb3IgT1M9PSJhbmRyb2lkIicsIHsK KyAgICAgICAgWyd1c2VfeDExPT0xJywgeworICAgICAgICAgICdkZXBlbmRlbmNpZXMnOiBbCisg ICAgICAgICAgICAnPChjaHJvbWl1bV9zcmNfZGlyKS90aGlyZF9wYXJ0eS9oYXJmYnV6ei1uZy9o YXJmYnV6ei5neXA6aGFyZmJ1enotbmcnLAorICAgICAgICAgIF0sCisgICAgICAgIH1dLAorICAg ICAgICBbJ09TPT0iYW5kcm9pZCInLCB7CiAgICAgICAgICAgJ2RlcGVuZGVuY2llcyc6IFsKICAg ICAgICAgICAgICc8KGNocm9taXVtX3NyY19kaXIpL3RoaXJkX3BhcnR5L2hhcmZidXp6L2hhcmZi dXp6Lmd5cDpoYXJmYnV6eicsCiAgICAgICAgICAgXSwKZGlmZiAtLWdpdCBhL1NvdXJjZS9XZWJL aXQvY2hyb21pdW0vZmVhdHVyZXMuZ3lwaSBiL1NvdXJjZS9XZWJLaXQvY2hyb21pdW0vZmVhdHVy ZXMuZ3lwaQppbmRleCA0ZWZlYWJjMzBlYzIxODUwYjhmYWQ5YjZmNGY3ZDYxMTA2YWRmMzg4Li5j YjFlNjM3OTlkNWNiYzAwNjcyYTA3MWM0MTE1MDkyODUyZDE2MWRkIDEwMDY0NAotLS0gYS9Tb3Vy Y2UvV2ViS2l0L2Nocm9taXVtL2ZlYXR1cmVzLmd5cGkKKysrIGIvU291cmNlL1dlYktpdC9jaHJv bWl1bS9mZWF0dXJlcy5neXBpCkBAIC0xNDAsNyArMTQwLDYgQEAKICAgICAgICdlbmFibGVfdG91 Y2hfZXZlbnRzJSc6IDEsCiAgICAgICAnZW5hYmxlX3RvdWNoX2ljb25fbG9hZGluZyUnIDogMCwK ICAgICAgICdlbmFibGVfbXV0YXRpb25fb2JzZXJ2ZXJzJSc6IDEsCi0gICAgICAndXNlX2hhcmZi dXp6X25nJSc6IDAsCiAgICAgfSwKICAgICAndXNlX2FjY2VsZXJhdGVkX2NvbXBvc2l0aW5nJSc6 ICc8KHVzZV9hY2NlbGVyYXRlZF9jb21wb3NpdGluZyknLAogICAgICdlbmFibGVfc2tpYV90ZXh0 JSc6ICc8KGVuYWJsZV9za2lhX3RleHQpJywKQEAgLTIyMyw3ICsyMjIsNyBAQAogICAgICAgICAg ICdXVEZfVVNFX1NLSUFfT05fTUFDX0NIUk9NSVVNPTEnLAogICAgICAgICBdLAogICAgICAgfV0s Ci0gICAgICBbJ3VzZV9oYXJmYnV6el9uZz09MScsIHsKKyAgICAgIFsndXNlX3gxMT09MScsIHsK ICAgICAgICAgJ2ZlYXR1cmVfZGVmaW5lcyc6IFsKICAgICAgICAgICAnV1RGX1VTRV9IQVJGQlVa Wl9ORz0xJywKICAgICAgICAgXSwK 168313 2012-10-11 16:35:23 -0700 2012-10-11 20:36:30 -0700 Add temporary expectations bug-97281-20121012083419.patch text/plain 11992 bashi U3VidmVyc2lvbiBSZXZpc2lvbjogMTMxMDk4CmRpZmYgLS1naXQgYS9Tb3VyY2UvV2ViQ29yZS9D aGFuZ2VMb2cgYi9Tb3VyY2UvV2ViQ29yZS9DaGFuZ2VMb2cKaW5kZXggOTJmM2EyZDA5MTUyMTBl ZTRiMjY3MGUyOTNlZmRkY2NkMWM4NzZlYy4uMzM4YjRkYmE1NGQzMWU5NjRiYzJlY2MwNjNjYWIx YzIzMDU3NTk5ZCAxMDA2NDQKLS0tIGEvU291cmNlL1dlYkNvcmUvQ2hhbmdlTG9nCisrKyBiL1Nv dXJjZS9XZWJDb3JlL0NoYW5nZUxvZwpAQCAtMSwzICsxLDE3IEBACisyMDEyLTEwLTExICBLZW5p Y2hpIElzaGliYXNoaSAgPGJhc2hpQGNocm9taXVtLm9yZz4KKworICAgICAgICBbQ2hyb21pdW1d IFVzZSBoYXJmYnV6ei1uZyBieSBkZWZhdWx0IG9uIExpbnV4CisgICAgICAgIGh0dHBzOi8vYnVn cy53ZWJraXQub3JnL3Nob3dfYnVnLmNnaT9pZD05NzI4MQorCisgICAgICAgIFJldmlld2VkIGJ5 IE5PQk9EWSAoT09QUyEpLgorCisgICAgICAgIFN3aXRjaCBvbGQgaGFyZmJ1enogdG8gaGFyZmJ1 enotbmcgb24gY2hyb21pdW0gbGludXggcG9ydC4KKworICAgICAgICBObyBuZXcgdGVzdHMuIEV4 aXN0aW5nIHRlc3RzIHNob3VsZCBjb3ZlciB0aGlzIHN3aXRjaC4KKyAgICAgICAgVGVzdCBleHBl Y3RhdGlvbnMgd2lsbCBiZSB1cGRhdGVkLgorCisgICAgICAgICogV2ViQ29yZS5neXAvV2ViQ29y ZS5neXA6CisKIDIwMTItMTAtMTEgIEpvc2h1YSBCZWxsICA8anNiZWxsQGNocm9taXVtLm9yZz4K IAogICAgICAgICBJbmRleGVkREI6IElETCB0eXBlcyBkZWZpbmVkIGluIHNwZWMgc2hvdWxkIGJl IHZpc2libGUgdG8gc2NyaXB0cwpkaWZmIC0tZ2l0IGEvU291cmNlL1dlYktpdC9jaHJvbWl1bS9D aGFuZ2VMb2cgYi9Tb3VyY2UvV2ViS2l0L2Nocm9taXVtL0NoYW5nZUxvZwppbmRleCA4MDNmMDIz YjcwOWY4NWFlMzA1MTk0MGFlNGZjYjRjY2UzYWU1NWUxLi5jZGI4N2YzYWY4NWY4ODA3ZGEyOGVh M2ExMjg1YWE2ZWI0ODEyMzhmIDEwMDY0NAotLS0gYS9Tb3VyY2UvV2ViS2l0L2Nocm9taXVtL0No YW5nZUxvZworKysgYi9Tb3VyY2UvV2ViS2l0L2Nocm9taXVtL0NoYW5nZUxvZwpAQCAtMSwzICsx LDEyIEBACisyMDEyLTEwLTExICBLZW5pY2hpIElzaGliYXNoaSAgPGJhc2hpQGNocm9taXVtLm9y Zz4KKworICAgICAgICBbQ2hyb21pdW1dIFVzZSBoYXJmYnV6ei1uZyBieSBkZWZhdWx0IG9uIExp bnV4CisgICAgICAgIGh0dHBzOi8vYnVncy53ZWJraXQub3JnL3Nob3dfYnVnLmNnaT9pZD05NzI4 MQorCisgICAgICAgIFJldmlld2VkIGJ5IE5PQk9EWSAoT09QUyEpLgorCisgICAgICAgICogZmVh dHVyZXMuZ3lwaTogRW5hYmxlIFdURl9VU0VfSEFSRkJVWlpfTkcgd2hlbiB1c2VfeDExID09IDEu CisKIDIwMTItMTAtMTEgIEFkYW0gQmFydGggIDxhYmFydGhAd2Via2l0Lm9yZz4KIAogICAgICAg ICBbQ2hyb21pdW1dIEltcHJvdmUgdGhlIGNvbW1lbnQgZGVzY3JpYmluZyBtYWluV29ybGRTY3Jp cHRDb250ZXh0CmRpZmYgLS1naXQgYS9Tb3VyY2UvV2ViQ29yZS9XZWJDb3JlLmd5cC9XZWJDb3Jl Lmd5cCBiL1NvdXJjZS9XZWJDb3JlL1dlYkNvcmUuZ3lwL1dlYkNvcmUuZ3lwCmluZGV4IGFmYmIz MWVjM2E3YjhlY2JmNzNkNDU0NWYzYzRkZTQxODQ1ZGFkMWYuLmMyYTgwNzg5NTJmNzcwMDEyODEw ODU3YTZkM2MzYTIwNmRlYTU2NDEgMTAwNjQ0Ci0tLSBhL1NvdXJjZS9XZWJDb3JlL1dlYkNvcmUu Z3lwL1dlYkNvcmUuZ3lwCisrKyBiL1NvdXJjZS9XZWJDb3JlL1dlYkNvcmUuZ3lwL1dlYkNvcmUu Z3lwCkBAIC00OCw4ICs0OCw2IEBACiAKICAgICAnZW5hYmxlX3dleGl0X3RpbWVfZGVzdHJ1Y3Rv cnMnOiAxLAogCi0gICAgJ3VzZV9oYXJmYnV6el9uZyUnOiAwLAotCiAgICAgJ3dlYmNvcmVfaW5j bHVkZV9kaXJzJzogWwogICAgICAgJy4uLycsCiAgICAgICAnLi4vLi4nLApAQCAtMjQ4LDcgKzI0 Niw3IEBACiAgICAgICAgICAgJy4uL3BsYXRmb3JtL2dyYXBoaWNzL2hhcmZidXp6JywKICAgICAg ICAgXSwKICAgICAgIH1dLAotICAgICAgWyd1c2VfeDExPT0xIGFuZCB1c2VfaGFyZmJ1enpfbmc9 PTEnLCB7CisgICAgICBbJ3VzZV94MTE9PTEnLCB7CiAgICAgICAgICd3ZWJjb3JlX2luY2x1ZGVf ZGlycyc6IFsKICAgICAgICAgICAnLi4vcGxhdGZvcm0vZ3JhcGhpY3MvaGFyZmJ1enovbmcnLAog ICAgICAgICBdLApAQCAtMTY2MywxMSArMTY2MSwxMiBAQAogICAgICAgICBbJ3VzZV94MTEgPT0g MScsIHsKICAgICAgICAgICAnc291cmNlcy8nOiBbCiAgICAgICAgICAgICAjIENoZXJyeS1waWNr IGZpbGVzIGV4Y2x1ZGVkIGJ5IHRoZSBicm9hZGVyIHJlZ3VsYXIgZXhwcmVzc2lvbnMgYWJvdmUu Ci0gICAgICAgICAgICBbJ2luY2x1ZGUnLCAncGxhdGZvcm0vZ3JhcGhpY3MvaGFyZmJ1enovQ29t cGxleFRleHRDb250cm9sbGVySGFyZkJ1enpcXC5jcHAkJ10sCiAgICAgICAgICAgICBbJ2luY2x1 ZGUnLCAncGxhdGZvcm0vZ3JhcGhpY3MvaGFyZmJ1enovRm9udEhhcmZCdXp6XFwuY3BwJCddLAog ICAgICAgICAgICAgWydpbmNsdWRlJywgJ3BsYXRmb3JtL2dyYXBoaWNzL2hhcmZidXp6L0ZvbnRQ bGF0Zm9ybURhdGFIYXJmQnV6elxcLmNwcCQnXSwKLSAgICAgICAgICAgIFsnaW5jbHVkZScsICdw bGF0Zm9ybS9ncmFwaGljcy9oYXJmYnV6ei9IYXJmQnV6elNraWFcXC5jcHAkJ10sCiAgICAgICAg ICAgICBbJ2luY2x1ZGUnLCAncGxhdGZvcm0vZ3JhcGhpY3MvaGFyZmJ1enovSGFyZkJ1enpTaGFw ZXJCYXNlXFwuKGNwcHxoKSQnXSwKKyAgICAgICAgICAgIFsnaW5jbHVkZScsICdwbGF0Zm9ybS9n cmFwaGljcy9oYXJmYnV6ei9uZy9IYXJmQnV6ek5HRmFjZVxcLihjcHB8aCkkJ10sCisgICAgICAg ICAgICBbJ2luY2x1ZGUnLCAncGxhdGZvcm0vZ3JhcGhpY3MvaGFyZmJ1enovbmcvSGFyZkJ1enpO R0ZhY2VTa2lhXFwuY3BwJCddLAorICAgICAgICAgICAgWydpbmNsdWRlJywgJ3BsYXRmb3JtL2dy YXBoaWNzL2hhcmZidXp6L25nL0hhcmZCdXp6U2hhcGVyXFwuKGNwcHxoKSQnXSwKICAgICAgICAg ICAgIFsnaW5jbHVkZScsICdwbGF0Zm9ybS9ncmFwaGljcy9vcGVudHlwZS9PcGVuVHlwZVR5cGVz XFwuaCQnXSwKICAgICAgICAgICAgIFsnaW5jbHVkZScsICdwbGF0Zm9ybS9ncmFwaGljcy9vcGVu dHlwZS9PcGVuVHlwZVZlcnRpY2FsRGF0YVxcLihjcHB8aCkkJ10sCiAgICAgICAgICAgICBbJ2lu Y2x1ZGUnLCAncGxhdGZvcm0vZ3JhcGhpY3Mvc2tpYS9TaW1wbGVGb250RGF0YVNraWFcXC5jcHAk J10sCkBAIC0xNjc4LDE1ICsxNjc3LDYgQEAKICAgICAgICAgICAgIFsnZXhjbHVkZScsICdIYXJm YnV6elteL10rXFwuKGNwcHxoKSQnXSwKICAgICAgICAgICBdLAogICAgICAgICB9XSwKLSAgICAg ICAgWyd1c2VfeDExPT0xIGFuZCB1c2VfaGFyZmJ1enpfbmc9PTEnLCB7Ci0gICAgICAgICAgJ3Nv dXJjZXMvJzogWwotICAgICAgICAgICAgWydleGNsdWRlJywgJ3BsYXRmb3JtL2dyYXBoaWNzL2hh cmZidXp6L0NvbXBsZXhUZXh0Q29udHJvbGxlckhhcmZCdXp6XFwuY3BwJCddLAotICAgICAgICAg ICAgWydleGNsdWRlJywgJ3BsYXRmb3JtL2dyYXBoaWNzL2hhcmZidXp6L0hhcmZCdXp6U2tpYVxc LmNwcCQnXSwKLSAgICAgICAgICAgIFsnaW5jbHVkZScsICdwbGF0Zm9ybS9ncmFwaGljcy9oYXJm YnV6ei9uZy9IYXJmQnV6ek5HRmFjZVxcLihjcHB8aCkkJ10sCi0gICAgICAgICAgICBbJ2luY2x1 ZGUnLCAncGxhdGZvcm0vZ3JhcGhpY3MvaGFyZmJ1enovbmcvSGFyZkJ1enpOR0ZhY2VTa2lhXFwu Y3BwJCddLAotICAgICAgICAgICAgWydpbmNsdWRlJywgJ3BsYXRmb3JtL2dyYXBoaWNzL2hhcmZi dXp6L25nL0hhcmZCdXp6U2hhcGVyXFwuKGNwcHxoKSQnXSwKLSAgICAgICAgICBdLAotICAgICAg ICB9XSwKICAgICAgICAgWyd0b29sa2l0X3VzZXNfZ3RrID09IDEnLCB7CiAgICAgICAgICAgJ3Nv dXJjZXMvJzogWwogICAgICAgICAgICAgIyBDaGVycnktcGljayBmaWxlcyBleGNsdWRlZCBieSB0 aGUgYnJvYWRlciByZWd1bGFyIGV4cHJlc3Npb25zIGFib3ZlLgpAQCAtMTY5Nyw3ICsxNjg3LDEy IEBACiAgICAgICAgICAgICBbJ2V4Y2x1ZGUnLCAnR3RrXFwuY3BwJCddLAogICAgICAgICAgIF0s CiAgICAgICAgIH1dLAotICAgICAgICBbJ3VzZV94MTE9PTEgb3IgT1M9PSJhbmRyb2lkIicsIHsK KyAgICAgICAgWyd1c2VfeDExPT0xJywgeworICAgICAgICAgICdkZXBlbmRlbmNpZXMnOiBbCisg ICAgICAgICAgICAnPChjaHJvbWl1bV9zcmNfZGlyKS90aGlyZF9wYXJ0eS9oYXJmYnV6ei1uZy9o YXJmYnV6ei5neXA6aGFyZmJ1enotbmcnLAorICAgICAgICAgIF0sCisgICAgICAgIH1dLAorICAg ICAgICBbJ09TPT0iYW5kcm9pZCInLCB7CiAgICAgICAgICAgJ2RlcGVuZGVuY2llcyc6IFsKICAg ICAgICAgICAgICc8KGNocm9taXVtX3NyY19kaXIpL3RoaXJkX3BhcnR5L2hhcmZidXp6L2hhcmZi dXp6Lmd5cDpoYXJmYnV6eicsCiAgICAgICAgICAgXSwKZGlmZiAtLWdpdCBhL1NvdXJjZS9XZWJL aXQvY2hyb21pdW0vZmVhdHVyZXMuZ3lwaSBiL1NvdXJjZS9XZWJLaXQvY2hyb21pdW0vZmVhdHVy ZXMuZ3lwaQppbmRleCA0ZWZlYWJjMzBlYzIxODUwYjhmYWQ5YjZmNGY3ZDYxMTA2YWRmMzg4Li5j YjFlNjM3OTlkNWNiYzAwNjcyYTA3MWM0MTE1MDkyODUyZDE2MWRkIDEwMDY0NAotLS0gYS9Tb3Vy Y2UvV2ViS2l0L2Nocm9taXVtL2ZlYXR1cmVzLmd5cGkKKysrIGIvU291cmNlL1dlYktpdC9jaHJv bWl1bS9mZWF0dXJlcy5neXBpCkBAIC0xNDAsNyArMTQwLDYgQEAKICAgICAgICdlbmFibGVfdG91 Y2hfZXZlbnRzJSc6IDEsCiAgICAgICAnZW5hYmxlX3RvdWNoX2ljb25fbG9hZGluZyUnIDogMCwK ICAgICAgICdlbmFibGVfbXV0YXRpb25fb2JzZXJ2ZXJzJSc6IDEsCi0gICAgICAndXNlX2hhcmZi dXp6X25nJSc6IDAsCiAgICAgfSwKICAgICAndXNlX2FjY2VsZXJhdGVkX2NvbXBvc2l0aW5nJSc6 ICc8KHVzZV9hY2NlbGVyYXRlZF9jb21wb3NpdGluZyknLAogICAgICdlbmFibGVfc2tpYV90ZXh0 JSc6ICc8KGVuYWJsZV9za2lhX3RleHQpJywKQEAgLTIyMyw3ICsyMjIsNyBAQAogICAgICAgICAg ICdXVEZfVVNFX1NLSUFfT05fTUFDX0NIUk9NSVVNPTEnLAogICAgICAgICBdLAogICAgICAgfV0s Ci0gICAgICBbJ3VzZV9oYXJmYnV6el9uZz09MScsIHsKKyAgICAgIFsndXNlX3gxMT09MScsIHsK ICAgICAgICAgJ2ZlYXR1cmVfZGVmaW5lcyc6IFsKICAgICAgICAgICAnV1RGX1VTRV9IQVJGQlVa Wl9ORz0xJywKICAgICAgICAgXSwKZGlmZiAtLWdpdCBhL0xheW91dFRlc3RzL0NoYW5nZUxvZyBi L0xheW91dFRlc3RzL0NoYW5nZUxvZwppbmRleCBhNDgyOTA1Yjg4NmIyMmJmNzQ0Mzk4ODUwNjdi YWRiNjI2MGFkZjg5Li5kMGNiMjNmMDM2NzEzNGZiNDA0ODY4Mzk4Y2ZhNzAwZGVkODUxMDgxIDEw MDY0NAotLS0gYS9MYXlvdXRUZXN0cy9DaGFuZ2VMb2cKKysrIGIvTGF5b3V0VGVzdHMvQ2hhbmdl TG9nCkBAIC0xLDMgKzEsMTIgQEAKKzIwMTItMTAtMTEgIEtlbmljaGkgSXNoaWJhc2hpICA8YmFz aGlAY2hyb21pdW0ub3JnPgorCisgICAgICAgIFtDaHJvbWl1bV0gVXNlIGhhcmZidXp6LW5nIGJ5 IGRlZmF1bHQgb24gTGludXgKKyAgICAgICAgaHR0cHM6Ly9idWdzLndlYmtpdC5vcmcvc2hvd19i dWcuY2dpP2lkPTk3MjgxCisKKyAgICAgICAgUmV2aWV3ZWQgYnkgTk9CT0RZIChPT1BTISkuCisK KyAgICAgICAgKiBwbGF0Zm9ybS9jaHJvbWl1bS9UZXN0RXhwZWN0YXRpb25zOiBBZGRlZCB0ZW1w b3JhcnkgZXhwZWN0YXRpb25zLgorCiAyMDEyLTEwLTExICBBZGFtIEJhcnRoICA8YWJhcnRoQHdl YmtpdC5vcmc+CiAKICAgICAgICAgT25lIG1vcmUgYmFzZWxpbmUgdHdlYWsgYWZ0ZXIgdjgtaTE4 biByb2xsLgpkaWZmIC0tZ2l0IGEvTGF5b3V0VGVzdHMvcGxhdGZvcm0vY2hyb21pdW0vVGVzdEV4 cGVjdGF0aW9ucyBiL0xheW91dFRlc3RzL3BsYXRmb3JtL2Nocm9taXVtL1Rlc3RFeHBlY3RhdGlv bnMKaW5kZXggMjUxN2JiMmZhYTMxZTUwZTVmMGVhZjM5ODkxMDZkNDkwYjhmZWY3Mi4uNGYxYWIy Njc2ZDkwNDBmMGQ5NDIyYzQyM2JiM2U3OGMxNzEwNTkxOSAxMDA2NDQKLS0tIGEvTGF5b3V0VGVz dHMvcGxhdGZvcm0vY2hyb21pdW0vVGVzdEV4cGVjdGF0aW9ucworKysgYi9MYXlvdXRUZXN0cy9w bGF0Zm9ybS9jaHJvbWl1bS9UZXN0RXhwZWN0YXRpb25zCkBAIC0zOTEwLDggKzM5MTAsNjEgQEAg d2Via2l0Lm9yZy9iLzk2OTYyIGZhc3QvanMvY3Jvc3MtZnJhbWUtcmVhbGx5LWJhZC10aW1lLXdp dGgtX19wcm90b19fLmh0bWwgWyBGYWkKIAogd2Via2l0Lm9yZy9iLzk3MTc5IGh0dHAvdGVzdHMv Y3NzL2xpbmstY3NzLWRpc2FibGVkLXZhbHVlLXdpdGgtc2xvdy1sb2FkaW5nLXNoZWV0Lmh0bWwg WyBGYWlsdXJlIFBhc3MgXQogCi0jIFRoZSB0ZXN0IHNob3VsZCBwYXNzIGFmdGVyIGhhcmZidXp6 IHRyYW5zaXRpb24uCi13ZWJraXQub3JnL2IvOTcyOTYgWyBMaW51eCBXaW4gXSBmYXN0L3RleHQv aW50ZXJuYXRpb25hbC9oZWJyZXctc2VsZWN0aW9uLmh0bWwgWyBJbWFnZU9ubHlGYWlsdXJlIF0K K3dlYmtpdC5vcmcvYi85NzI5NiBbIFdpbiBdIGZhc3QvdGV4dC9pbnRlcm5hdGlvbmFsL2hlYnJl dy1zZWxlY3Rpb24uaHRtbCBbIEltYWdlT25seUZhaWx1cmUgXQorCisjIFRlbXBvcmFyeSBleHBl Y3RhdGlvbnMgZHVyaW5nIGhhcmZidXp6IHRyYW5zaXRpb24uIFdpbGwgYmUgcmVtb3ZlZCBhZnRl ciB0aGUgdHJhbnNpdGlvbiBpcyBjb21wbGV0ZWQuCit3ZWJraXQub3JnL2IvOTcyODEgWyBMaW51 eCBdIGNzczMvZm9udC1mZWF0dXJlLXNldHRpbmdzLXJlbmRlcmluZy5odG1sIFsgRmFpbHVyZSBd Cit3ZWJraXQub3JnL2IvOTcyODEgWyBMaW51eCBdIGZhc3QvdGV4dC9hdHN1aS1uZWdhdGl2ZS1z cGFjaW5nLWZlYXR1cmVzLmh0bWwgWyBJbWFnZU9ubHlGYWlsdXJlIF0KK3dlYmtpdC5vcmcvYi85 NzI4MSBbIExpbnV4IF0gZmFzdC90ZXh0L2F0c3VpLXNwYWNpbmctZmVhdHVyZXMuaHRtbCBbIElt YWdlT25seUZhaWx1cmUgXQord2Via2l0Lm9yZy9iLzk3MjgxIFsgTGludXggXSBmYXN0L3RleHQv Y2ctdnMtYXRzdWkuaHRtbCBbIEltYWdlT25seUZhaWx1cmUgXQord2Via2l0Lm9yZy9iLzk3Mjgx IFsgTGludXggXSBmYXN0L3RleHQvY29tcGxleC10ZXh0LW9wYWNpdHkuaHRtbCBbIEltYWdlT25s eUZhaWx1cmUgXQord2Via2l0Lm9yZy9iLzk3MjgxIFsgTGludXggXSBmYXN0L3RleHQvZHJhd0Jp ZGlUZXh0Lmh0bWwgWyBJbWFnZU9ubHlGYWlsdXJlIF0KK3dlYmtpdC5vcmcvYi85NzI4MSBbIExp bnV4IF0gZmFzdC90ZXh0L2ludGVybmF0aW9uYWwvYXJhYmljLWp1c3RpZnkuaHRtbCBbIEltYWdl T25seUZhaWx1cmUgXQord2Via2l0Lm9yZy9iLzk3MjgxIFsgTGludXggXSBmYXN0L3RleHQvaW50 ZXJuYXRpb25hbC9oZWJyZXctdm93ZWxzLmh0bWwgWyBJbWFnZU9ubHlGYWlsdXJlIF0KK3dlYmtp dC5vcmcvYi85NzI4MSBbIExpbnV4IF0gZmFzdC90ZXh0L2ludGVybmF0aW9uYWwvaGluZGktc3Bh Y2luZy5odG1sIFsgSW1hZ2VPbmx5RmFpbHVyZSBdCit3ZWJraXQub3JnL2IvOTcyODEgWyBMaW51 eCBdIGZhc3QvdGV4dC9pbnRlcm5hdGlvbmFsL3RleHQtc3BsaWNlZC1mb250Lmh0bWwgWyBJbWFn ZU9ubHlGYWlsdXJlIF0KK3dlYmtpdC5vcmcvYi85NzI4MSBbIExpbnV4IF0gZmFzdC90ZXh0L2lu dGVybmF0aW9uYWwvdGhhaS1saW5lLWJyZWFrcy5odG1sIFsgSW1hZ2VPbmx5RmFpbHVyZSBdCit3 ZWJraXQub3JnL2IvOTcyODEgWyBMaW51eCBdIHBsYXRmb3JtL2Nocm9taXVtLWxpbnV4L2Zhc3Qv dGV4dC9pbnRlcm5hdGlvbmFsL2FyYWJpYy12ZXJ0aWNhbC1vZmZzZXQuaHRtbCBbIEltYWdlT25s eUZhaWx1cmUgXQord2Via2l0Lm9yZy9iLzk3MjgxIFsgTGludXggXSBwbGF0Zm9ybS9jaHJvbWl1 bS1saW51eC9mYXN0L3RleHQvaW50ZXJuYXRpb25hbC9jb21wbGV4LWpvaW5pbmctdXNpbmctZ3Bv cy5odG1sIFsgSW1hZ2VPbmx5RmFpbHVyZSBdCit3ZWJraXQub3JnL2IvOTcyODEgWyBMaW51eCBd IHBsYXRmb3JtL2Nocm9taXVtLWxpbnV4L3N2Zy90ZXh0L3RleHQtd2l0aC1nZW9tZXRyaWMtcHJl Y2lzaW9uLnN2ZyBbIEltYWdlT25seUZhaWx1cmUgXQord2Via2l0Lm9yZy9iLzk3MjgxIFsgTGlu dXggXSBzdmcvVzNDLUkxOE4vZy1kaXJMVFItdWJPdmVycmlkZS5zdmcgWyBJbWFnZU9ubHlGYWls dXJlIF0KK3dlYmtpdC5vcmcvYi85NzI4MSBbIExpbnV4IF0gc3ZnL1czQy1JMThOL2ctZGlyUlRM LXViT3ZlcnJpZGUuc3ZnIFsgSW1hZ2VPbmx5RmFpbHVyZSBdCit3ZWJraXQub3JnL2IvOTcyODEg WyBMaW51eCBdIHN2Zy9XM0MtSTE4Ti90ZXh0LWFuY2hvci1kaXJMVFItYW5jaG9yRW5kLnN2ZyBb IEltYWdlT25seUZhaWx1cmUgXQord2Via2l0Lm9yZy9iLzk3MjgxIFsgTGludXggXSBzdmcvVzND LUkxOE4vdGV4dC1hbmNob3ItZGlyTFRSLWFuY2hvck1pZGRsZS5zdmcgWyBJbWFnZU9ubHlGYWls dXJlIF0KK3dlYmtpdC5vcmcvYi85NzI4MSBbIExpbnV4IF0gc3ZnL1czQy1JMThOL3RleHQtYW5j aG9yLWRpckxUUi1hbmNob3JTdGFydC5zdmcgWyBJbWFnZU9ubHlGYWlsdXJlIF0KK3dlYmtpdC5v cmcvYi85NzI4MSBbIExpbnV4IF0gc3ZnL1czQy1JMThOL3RleHQtYW5jaG9yLWRpck5vbmUtYW5j aG9yRW5kLnN2ZyBbIEltYWdlT25seUZhaWx1cmUgXQord2Via2l0Lm9yZy9iLzk3MjgxIFsgTGlu dXggXSBzdmcvVzNDLUkxOE4vdGV4dC1hbmNob3ItZGlyTm9uZS1hbmNob3JNaWRkbGUuc3ZnIFsg SW1hZ2VPbmx5RmFpbHVyZSBdCit3ZWJraXQub3JnL2IvOTcyODEgWyBMaW51eCBdIHN2Zy9XM0Mt STE4Ti90ZXh0LWFuY2hvci1kaXJOb25lLWFuY2hvclN0YXJ0LnN2ZyBbIEltYWdlT25seUZhaWx1 cmUgXQord2Via2l0Lm9yZy9iLzk3MjgxIFsgTGludXggXSBzdmcvVzNDLUkxOE4vdGV4dC1hbmNo b3ItZGlyUlRMLWFuY2hvckVuZC5zdmcgWyBJbWFnZU9ubHlGYWlsdXJlIF0KK3dlYmtpdC5vcmcv Yi85NzI4MSBbIExpbnV4IF0gc3ZnL1czQy1JMThOL3RleHQtYW5jaG9yLWRpclJUTC1hbmNob3JN aWRkbGUuc3ZnIFsgSW1hZ2VPbmx5RmFpbHVyZSBdCit3ZWJraXQub3JnL2IvOTcyODEgWyBMaW51 eCBdIHN2Zy9XM0MtSTE4Ti90ZXh0LWFuY2hvci1kaXJSVEwtYW5jaG9yU3RhcnQuc3ZnIFsgSW1h Z2VPbmx5RmFpbHVyZSBdCit3ZWJraXQub3JnL2IvOTcyODEgWyBMaW51eCBdIHN2Zy9XM0MtSTE4 Ti90ZXh0LWFuY2hvci1pbmhlcml0ZWQtZGlyTFRSLWFuY2hvckVuZC5zdmcgWyBJbWFnZU9ubHlG YWlsdXJlIF0KK3dlYmtpdC5vcmcvYi85NzI4MSBbIExpbnV4IF0gc3ZnL1czQy1JMThOL3RleHQt YW5jaG9yLWluaGVyaXRlZC1kaXJMVFItYW5jaG9yTWlkZGxlLnN2ZyBbIEltYWdlT25seUZhaWx1 cmUgXQord2Via2l0Lm9yZy9iLzk3MjgxIFsgTGludXggXSBzdmcvVzNDLUkxOE4vdGV4dC1hbmNo b3ItaW5oZXJpdGVkLWRpckxUUi1hbmNob3JTdGFydC5zdmcgWyBJbWFnZU9ubHlGYWlsdXJlIF0K K3dlYmtpdC5vcmcvYi85NzI4MSBbIExpbnV4IF0gc3ZnL1czQy1JMThOL3RleHQtYW5jaG9yLWlu aGVyaXRlZC1kaXJSVEwtYW5jaG9yRW5kLnN2ZyBbIEltYWdlT25seUZhaWx1cmUgXQord2Via2l0 Lm9yZy9iLzk3MjgxIFsgTGludXggXSBzdmcvVzNDLUkxOE4vdGV4dC1hbmNob3ItaW5oZXJpdGVk LWRpclJUTC1hbmNob3JNaWRkbGUuc3ZnIFsgSW1hZ2VPbmx5RmFpbHVyZSBdCit3ZWJraXQub3Jn L2IvOTcyODEgWyBMaW51eCBdIHN2Zy9XM0MtSTE4Ti90ZXh0LWFuY2hvci1pbmhlcml0ZWQtZGly UlRMLWFuY2hvclN0YXJ0LnN2ZyBbIEltYWdlT25seUZhaWx1cmUgXQord2Via2l0Lm9yZy9iLzk3 MjgxIFsgTGludXggXSBzdmcvVzNDLUkxOE4vdGV4dC1hbmNob3Itbm8tbWFya3VwLnN2ZyBbIElt YWdlT25seUZhaWx1cmUgXQord2Via2l0Lm9yZy9iLzk3MjgxIFsgTGludXggXSBzdmcvVzNDLUkx OE4vdGV4dC1kaXJSVEwtdWJPdmVycmlkZS5zdmcgWyBJbWFnZU9ubHlGYWlsdXJlIF0KK3dlYmtp dC5vcmcvYi85NzI4MSBbIExpbnV4IF0gc3ZnL1czQy1JMThOL3RzcGFuLWRpckxUUi11YkVtYmVk LWluLXJ0bC1jb250ZXh0LnN2ZyBbIEltYWdlT25seUZhaWx1cmUgXQord2Via2l0Lm9yZy9iLzk3 MjgxIFsgTGludXggXSBzdmcvVzNDLUkxOE4vdHNwYW4tZGlyTFRSLXViTm9uZS1pbi1ydGwtY29u dGV4dC5zdmcgWyBJbWFnZU9ubHlGYWlsdXJlIF0KK3dlYmtpdC5vcmcvYi85NzI4MSBbIExpbnV4 IF0gc3ZnL1czQy1JMThOL3RzcGFuLWRpckxUUi11Yk92ZXJyaWRlLWluLXJ0bC1jb250ZXh0LnN2 ZyBbIEltYWdlT25seUZhaWx1cmUgXQord2Via2l0Lm9yZy9iLzk3MjgxIFsgTGludXggXSBzdmcv VzNDLUkxOE4vdHNwYW4tZGlyTm9uZS11Yk92ZXJyaWRlLWluLXJ0bC1jb250ZXh0LnN2ZyBbIElt YWdlT25seUZhaWx1cmUgXQord2Via2l0Lm9yZy9iLzk3MjgxIFsgTGludXggXSBzdmcvVzNDLUkx OE4vdHNwYW4tZGlyUlRMLXViT3ZlcnJpZGUtaW4tZGVmYXVsdC1jb250ZXh0LnN2ZyBbIEltYWdl T25seUZhaWx1cmUgXQord2Via2l0Lm9yZy9iLzk3MjgxIFsgTGludXggXSBzdmcvVzNDLUkxOE4v dHNwYW4tZGlyUlRMLXViT3ZlcnJpZGUtaW4tbHRyLWNvbnRleHQuc3ZnIFsgSW1hZ2VPbmx5RmFp bHVyZSBdCit3ZWJraXQub3JnL2IvOTcyODEgWyBMaW51eCBdIHN2Zy9XM0MtSTE4Ti90c3Bhbi1k aXJlY3Rpb24tcnRsLnN2ZyBbIEltYWdlT25seUZhaWx1cmUgXQord2Via2l0Lm9yZy9iLzk3Mjgx IFsgTGludXggXSBzdmcvVzNDLVNWRy0xLjEtU0UvdGV4dC1pbnRyby0wNS10LnN2ZyBbIEltYWdl T25seUZhaWx1cmUgXQord2Via2l0Lm9yZy9iLzk3MjgxIFsgTGludXggXSBzdmcvVzNDLVNWRy0x LjEvdGV4dC1pbnRyby0wNS10LnN2ZyBbIEltYWdlT25seUZhaWx1cmUgXQord2Via2l0Lm9yZy9i Lzk3MjgxIFsgTGludXggXSBzdmcvdGV4dC9iaWRpLXRzcGFucy5zdmcgWyBJbWFnZU9ubHlGYWls dXJlIF0KK3dlYmtpdC5vcmcvYi85NzI4MSBbIExpbnV4IF0gc3ZnL3RleHQvdGV4dC1pbnRyby0w NS10LnN2ZyBbIEltYWdlT25seUZhaWx1cmUgXQord2Via2l0Lm9yZy9iLzk3MjgxIFsgTGludXgg XSB0cmFuc2Zvcm1zLzJkL2hpbmRpLXJvdGF0ZWQuaHRtbCBbIEltYWdlT25seUZhaWx1cmUgXQor d2Via2l0Lm9yZy9iLzk3MjgxIFsgTGludXggXSBmYXN0L3RleHQvaW50ZXJuYXRpb25hbC9raG1l ci1zZWxlY3Rpb24uaHRtbCBbIEZhaWx1cmUgXQord2Via2l0Lm9yZy9iLzk3MjgxIFsgTGludXgg XSBzdmcvVzNDLUkxOE4vdGV4dC1kaXJMVFItdWJPdmVycmlkZS5zdmcgWyBGYWlsdXJlIF0KK3dl YmtpdC5vcmcvYi85NzI4MSBbIExpbnV4IF0gc3ZnL1czQy1JMThOL3RzcGFuLWRpckxUUi11Yk92 ZXJyaWRlLWluLWRlZmF1bHQtY29udGV4dC5zdmcgWyBGYWlsdXJlIF0KK3dlYmtpdC5vcmcvYi85 NzI4MSBbIExpbnV4IF0gc3ZnL1czQy1JMThOL3RzcGFuLWRpckxUUi11Yk92ZXJyaWRlLWluLWx0 ci1jb250ZXh0LnN2ZyBbIEZhaWx1cmUgXQord2Via2l0Lm9yZy9iLzk3MjgxIFsgTGludXggXSBz dmcvVzNDLUkxOE4vdHNwYW4tZGlyTm9uZS11Yk92ZXJyaWRlLWluLWRlZmF1bHQtY29udGV4dC5z dmcgWyBGYWlsdXJlIF0KK3dlYmtpdC5vcmcvYi85NzI4MSBbIExpbnV4IF0gc3ZnL1czQy1JMThO L3RzcGFuLWRpck5vbmUtdWJPdmVycmlkZS1pbi1sdHItY29udGV4dC5zdmcgWyBGYWlsdXJlIF0K K3dlYmtpdC5vcmcvYi85NzI4MSBbIExpbnV4IF0gc3ZnL3RleHQvYmlkaS10ZXh0LXF1ZXJ5LnN2 ZyBbIEZhaWx1cmUgXQogCiAjIEV4cGVjdGVkIHRvIGZhaWwgdW50aWwgc3RvcmFnZSBibG9ja2lu ZyBzZXR0aW5ncyBoYXZlIGEgV2ViS2l0MSBBUEkgaW4gQ2hyb21pdW0uCiB3ZWJraXQub3JnL2Iv OTc0OTEgaHR0cC90ZXN0cy9zZWN1cml0eS9jcm9zcy1vcmlnaW4tbG9jYWwtc3RvcmFnZS13azEu aHRtbCBbIEZhaWx1cmUgUGFzcyBdCg==