41813 2010-07-07 16:59:04 -0700 XHRs get unnecessary Cache-Control http header after reload 2010-07-13 04:32:01 -0700 1 1 1 Unclassified WebKit WebCore Misc. 528+ (Nightly build) PC OS X 10.5 RESOLVED FIXED P2 Normal --- 1 japhet japhet ap beidson fishd koivisto mario mjs oldest_to_newest 247859 0 japhet 2010-07-07 16:59:04 -0700 When a page is reloaded, it verifies all items in the cache with the server instead of just loading from cache if permitted. The state that controls this action is not cleared when the load is completed (it is cleared during the next navigation), so any subsequent loads (e.g., XHRs) will be given a Cache-Control header, even if the reload "action" has completed. I believe the state should be cleared after the load event rather than always waiting until the next navigation. (This causes a lot of extra http requests that get 304 responses on sites like Facebook, where a bunch of static content is loaded via async XHR after the "page has loaded". Initial report at http://code.google.com/p/chromium/issues/detail?id=7819) 247865 1 60807 japhet 2010-07-07 17:08:38 -0700 Created attachment 60807 Patch 248029 2 60807 fishd 2010-07-07 23:13:06 -0700 Comment on attachment 60807 Patch WebCore/loader/FrameLoader.cpp:2606 > + m_loadType = FrameLoadTypeStandard; I think we also want to reset the load type to standard after performing a back/forward navigation. Are there any cases when we wouldn't want to reset it? WebCore/loader/SubresourceLoader.cpp:96 > + else if (!fl->activeDocumentLoader()->isLoadingInAPISense()) I would expect subresources to only care about the state of the documentLoader. It seems wrong to care if there is a new provisional navigation starting. Are both of these changes necessary? It seems like the change to SubresourceLoader.cpp should be sufficient. However, resetting m_loadType certainly seems reasonable to me. 248030 3 60807 fishd 2010-07-07 23:13:50 -0700 Comment on attachment 60807 Patch R- because of the activeDocumentLoader thing. 248291 4 ap 2010-07-08 10:01:39 -0700 This seems intentional. What's the difference between content loaded before onload and after onload? If a user initiates a reload, they want to refresh content, and technical details of whether it's loaded via HTML tags or XMLHttpRequest are irrelevant. In fact, many people complain about the opposite behavior - we sometimes DON'T revalidate resources after reload, see bug 30862 and related. 248328 5 fishd 2010-07-08 10:47:28 -0700 (In reply to comment #4) > This seems intentional. This behavior differs from Mozilla and IE, and it means that WebKit based browsers generate nearly 3x as many network requests as those browsers on Facebook! Any site that navigates using reference fragments (or history.pushState) to reuse the existing document will perform poorly at the network level after the user hits reload once on a particular "page". > What's the difference between content loaded before > onload and after onload? Content loaded during page load is related to the page load. You can see this concept applied to other aspects of page loading: a subframe navigated during page load does not create a separate back/forward entry. Consider back/forward navigations. When navigating back in history, we apply a rule that allows stale content to be loaded from the cache. However, if the user then interacts with the page, and the page creates an XHR, should that XHR also inherit the rule that causes it to load stale content? It seems like that could render a web app unusable. I think the load flags should be temporary for the duration of a page load (navigation). After that, they should reset to the normal state, so that future interactions will behave as they do when a page is normally loaded. > If a user initiates a reload, they want to refresh > content, and technical details of whether it's loaded via HTML tags or > XMLHttpRequest are irrelevant. Agreed. Nate's patch treats all subresource loads equally. > In fact, many people complain about the opposite behavior - we sometimes > DON'T revalidate resources after reload, see bug 30862 and related. Bug 30862 is not contradictory to this bug. The script element created during parsing of the body should inherit the load flags used to fetch the document. The load event should be deferred until that script element's load event handler runs, but during that event handler if another script element is created, then the document's load event will be further deferred, and the second script element should also inherit the reload flag. Finally, when the second script element finishes loading, we can fire the load event to the page and reset the load flags. No problem. 248354 6 ap 2010-07-08 11:15:42 -0700 Matching other browsers is of course a very strong argument. However, we've been long trying to apply our own thinking to refresh functionality, and didn't just blindly follow others here. > Agreed. Nate's patch treats all subresource loads equally. It's only equal in a low level technical sense. The user doesn't know which requests are made before onload, and which are made after it. Actually, it's dynamic content that's most interesting to refresh for normal people, so preventing revalidation in that case sounds like a bad thing. 248375 7 fishd 2010-07-08 11:36:03 -0700 (In reply to comment #6) > Matching other browsers is of course a very strong argument. However, we've > been long trying to apply our own thinking to refresh functionality, and didn't > just blindly follow others here. 3x the number of network requests on Facebook is not in the realm of reasonable, don't you agree? I mean, it is really bad. There are a lot of Facebook users. I agree with the principle of doing what's best provided it is compatible with the web. In this case, I think the choice is pretty clear. > > Agreed. Nate's patch treats all subresource loads equally. > > It's only equal in a low level technical sense. The user doesn't know which > requests are made before onload, and which are made after it. Actually, it's > dynamic content that's most interesting to refresh for normal people, so > preventing revalidation in that case sounds like a bad thing. I guess I was confused by your argument. I agree that the user won't know the difference for requests made close to the time of the page load stopping. However, after the page has been quiescent for a little while, I don't think the user should expect a previous reload to impact new clicks. I think there is some burden on the web page developer to do the right thing when using XHR. If we apply load flags to any request made before or during onload, then the web developer can make sure to issue subresource requests during that time so that a 'reload' would apply to those requests. This will work in Firefox and IE (see bug 30862). Web developers can similarly know that load flags will not apply to requests issued after onload. So, they can plan accordingly. I know there is still the case of uninformed web developers, and that is surely the majority case, but I don't know what else to do. The current behavior is hurting more than it is helping. I think that's clear. 248378 8 fishd 2010-07-08 11:38:05 -0700 (In reply to comment #7) > (In reply to comment #6) > > Matching other browsers is of course a very strong argument. However, we've > > been long trying to apply our own thinking to refresh functionality, and didn't > > just blindly follow others here. > > 3x the number of network requests on Facebook is not in the realm of > reasonable, don't you agree? In case you were wondering, this issue was reported by Facebook developers. The 3x number comes from their network traffic logs. 248390 9 ap 2010-07-08 11:51:34 -0700 Is it known why users try to reload Facebook pages? Will this change make their lives worse? > However, after the page has been quiescent for a little while, I don't think > the user should expect a previous reload to impact new clicks. I hesitate to propose it, but this almost sounds like there should be a timeout of several seconds then. 248443 10 fishd 2010-07-08 12:45:40 -0700 (In reply to comment #9) > Is it known why users try to reload Facebook pages? Will this change make their > lives worse? I suspect people hit the reload button to reload the page they are looking at. I doubt they do so to reload the page that they will look at next. You know what I mean? On facebook, there is only one real page, but to the user it looks like many pages. The user probably only wants to reload a single conceptual page, which is what this patch will give them. That's my intuition anyways. > > However, after the page has been quiescent for a little while, I don't think > > the user should expect a previous reload to impact new clicks. > > I hesitate to propose it, but this almost sounds like there should be a timeout > of several seconds then. Let's start by having a hard cut off after onload, and fix bug 30862 as I proposed. I bet that will address most user complaints. Do you know of any other bugs related to this topic besides bug 30862? 248494 11 ap 2010-07-08 13:36:17 -0700 > I suspect people hit the reload button to reload the page they are looking at. > I doubt they do so to reload the page that they will look at next. Well, this just means that we should reset the load type when fragment changes, doesn't it? My question was a bit different though. What kind of content do people want to refresh on Facebook? It's possible that fixing this bug in this way will expose some other difference in behavior, and WebKit will be the only engine that doesn't refresh this content. That concern is of course not limited to Facebook, but we could start testing with it if we knew what scenario we're talking about. > Let's start by having a hard cut off after onload, and fix bug 30862 as I proposed. Fixing 30862 would be of course very welcome. I'd love to hear what other folks CC'ed here think about your proposal about this bug. > Do you know of any other bugs related to this topic besides bug 30862? There are several bugs mentioned in comments there. Besides those, bug 38690 may be related (it might even be related to Facebook problem - do we know for sure if it's users actually hitting Refresh, and not some other reason that makes WebKit revalidate?) 248530 12 koivisto 2010-07-08 14:11:16 -0700 I agree with Darin about both this and 30862. The current behavior is buggy and resetting the load type on onload is the appropriate fix. I had noticed the problem earlier when working with the loader code but never tried to fix it as there wasn't known serious harm caused by it (reloads are assumed to be relatively rare) and the fix is a risky behavior change. Since it is now known to cause real pain, the fix should be tried out. 248536 13 fishd 2010-07-08 14:18:03 -0700 (In reply to comment #11) > > I suspect people hit the reload button to reload the page they are looking at. > > I doubt they do so to reload the page that they will look at next. > > Well, this just means that we should reset the load type when fragment changes, > doesn't it? That is another choice. It seems worse to me. Again, a request issued after page load completes seems conceptually divorced from page load. Once the page loads, the page should now act like a page that was loaded normally. > My question was a bit different though. What kind of content do people want > to refresh on Facebook? I don't know... maybe to see if people have posted additional comments? > It's possible that fixing this bug in this way will > expose some other difference in behavior, and WebKit will be the only > engine that doesn't refresh this content. The patch here is trying to make WebKit be more like other browsers. > That concern is of course not > limited to Facebook, but we could start testing with it if we knew what > scenario we're talking about. The scenario is given in comment #0. Use XHR to request a resource. > > Do you know of any other bugs related to this topic besides bug 30862? > > There are several bugs mentioned in comments there. Besides those, bug 38690 > may be related (it might even be related to Facebook problem - do we know for > sure if it's users actually hitting Refresh, and not some other reason that > makes WebKit revalidate?) I don't know for sure. It is a guess based on observing the behavior of reload on Facebook pages. I don't observe the problem on Facebook unless I hit reload first. I didn't exhaust all possible avenues to see if there isn't a redirect following a form submission somewhere. It is possible, but given the popularity of the reload button, I'm tempted to blame that. 248558 14 ap 2010-07-08 14:40:52 -0700 I still see this change as likely to degrade user experience, but I won't argue with both you and Antti any more. >> It's possible that fixing this bug in this way will >> expose some other difference in behavior, and WebKit will be the only >> engine that doesn't refresh this content. > > The patch here is trying to make WebKit be more like other browsers. It's very common that a patch that technically makes us more like other browsers actually breaks functionality if we're still insufficiently like them (or if there are browser specific code paths). All I'm asking for is to actually test (to a reasonable degree) the use case we're supposed to be improving here. Not in the terms of XMLHttpRequest or DOM events, but in the terms of still seeing new content on Facebook. 248615 15 fishd 2010-07-08 16:16:07 -0700 (In reply to comment #14) > > The patch here is trying to make WebKit be more like other browsers. > > It's very common that a patch that technically makes us more like other > browsers actually breaks functionality if we're still insufficiently like them > (or if there are browser specific code paths). Yes, this is a good thing to be concerned about. > All I'm asking for is to > actually test (to a reasonable degree) the use case we're supposed to be > improving here. Not in the terms of XMLHttpRequest or DOM events, but in the > terms of still seeing new content on Facebook. Certainly. When we have this fix in the dev channel of Chrome, I plan on informing the Facebook engineers so they can also confirm the fix. 249290 16 61100 japhet 2010-07-09 14:41:34 -0700 Created attachment 61100 patch #2 This change merges the places that are setting cache policy based on DocumentLoader's original request into addExtraFieldsToRequest. As it currently stands, those settings often get overridden anyway, and it makes it much harder to follow the logic of the cache policy. Hopefully this is clearer. 249359 17 61100 fishd 2010-07-09 17:08:01 -0700 Comment on attachment 61100 patch #2 LayoutTests/http/tests/xmlhttprequest/cache-headers-after-reload.html:25 + xhr.onload = finish; nit: indentation R=me 249361 18 fishd 2010-07-09 17:11:28 -0700 By the way, Alexey... it seems we were already not observing this bug for subresources that go through the WebCore cache. Take a look at FrameLoader::subresourceCachePolicy. It will return early if m_isComplete is true, and that is true once we fire onload. So, Nate's patch has the effect of making subresources like XHR that do not go through the WebCore cache behave consistently with those that do with regards to reload behavior. We seriously need to clean up this code. It is really wonky that there are two systems determining cache policy for different subresources. 249960 19 japhet 2010-07-12 12:25:27 -0700 http://trac.webkit.org/changeset/63100 250162 20 mario 2010-07-12 17:08:33 -0700 (In reply to comment #19) > http://trac.webkit.org/changeset/63100 Not sure of the actual reason but it seems this patch is causing an API test failing in the GTK port: http://build.webkit.org/builders/GTK%20Linux%2032-bit%20Release/builds/1381/steps/API%20tests/logs/stdio I found this one as the guilty commit by using git bisect with latest code from trunk, but still don't know whether this test is failing because this test broke something or because it uncovered some other problem somwhere else. See comment #37 on bug 38648 for more details, now it's time for me to go to bed. 250175 21 mario 2010-07-12 17:25:10 -0700 (In reply to comment #20) > (In reply to comment #19) > > http://trac.webkit.org/changeset/63100 > > Not sure of the actual reason but it seems this patch is causing an API test failing in the GTK port: > > http://build.webkit.org/builders/GTK%20Linux%2032-bit%20Release/builds/1381/steps/API%20tests/logs/stdio > > I found this one as the guilty commit by using git bisect with latest code from trunk, but still don't know whether this test is failing because this test broke something or because it uncovered some other problem somwhere else. > > See comment #37 on bug 38648 for more details, now it's time for me to go to bed. Hmmm.. looks like I was too slow, as Martin already filed a bug for this: https://bugs.webkit.org/show_bug.cgi?id=42114 Sorry for the spam 250436 22 mario 2010-07-13 04:32:01 -0700 (In reply to comment #20) > (In reply to comment #19) > > http://trac.webkit.org/changeset/63100 > > Not sure of the actual reason but it seems this patch is causing an API test > failing in the GTK port: > > http://build.webkit.org/builders/GTK%20Linux%2032-bit%20Release/builds/1381/steps/API%20tests/logs/stdio > > I found this one as the guilty commit by using git bisect with latest code from > trunk, but still don't know whether this test is failing because this test > broke something or because it uncovered some other problem somwhere else. After taking another look on this (I'm feel really curious about this stuff), I found the problem in the WebKitGTK API test comes because of this one-line change: diff --git a/WebCore/loader/FrameLoader.cpp b/WebCore/loader/FrameLoader.cpp index 8611ac0..40beec7 100644 --- a/WebCore/loader/FrameLoader.cpp +++ b/WebCore/loader/FrameLoader.cpp @@ -2605,6 +2605,9 @@ String FrameLoader::userAgent(const KURL& url) const void FrameLoader::handledOnloadEvents() { m_client->dispatchDidHandleOnloadEvents(); + + m_loadType = FrameLoadTypeStandard; + #if ENABLE(OFFLINE_WEB_APPLICATIONS) if (documentLoader()) documentLoader()->applicationCacheHost()->stopDeferringEvents(); I'm not an expert on this matter so I can't tell why this is causing the following assertion in the webview unit test to fail: ERROR:../../WebKit/gtk/tests/testwebview.c:269:do_test_webkit_web_view_adjustments: assertion failed (gtk_adjustment_get_value(adjustment) == 100.0): (0 == 100) ...but somehow that's causing the problem and I wonder, out of my ignorance about this part of the code, whether the handleOnloadEvents() function is the right place to reset the m_loadTypeValue. Maybe with this piece of information someone of you could help to understand why this is happening and therefore to fix bug 42114. Thanks in advance. 60807 2010-07-07 17:08:38 -0700 2010-07-09 14:41:34 -0700 Patch xhr-cache.txt text/plain 6902 japhet SW5kZXg6IFdlYkNvcmUvQ2hhbmdlTG9nCj09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT0KLS0tIFdlYkNvcmUvQ2hhbmdlTG9n CShyZXZpc2lvbiA2MjcyOSkKKysrIFdlYkNvcmUvQ2hhbmdlTG9nCSh3b3JraW5nIGNvcHkpCkBA IC0xLDMgKzEsMjUgQEAKKzIwMTAtMDctMDcgIE5hdGUgQ2hhcGluICA8amFwaGV0QGNocm9taXVt Lm9yZz4KKworICAgICAgICBSZXZpZXdlZCBieSBOT0JPRFkgKE9PUFMhKS4KKworICAgICAgICBF bnN1cmUgdGhhdCBhIGNhY2hlIHBvbGljeSB0aGF0IGZvcmNlcyB2YWxpZGF0aW9uIGlzIGNsZWFy ZWQgb25jZQorICAgICAgICB0aGUgbG9hZCBldmVudCBpcyBmaXJlZCwgcmF0aGVyIHRoYW4gb25s eSBkb2luZyBzbyBhdCB0aGUgbmV4dAorICAgICAgICBuYXZpZ2F0aW9uLiBUaGlzIGxlYWRzIHRv IGEgbG90IG9mIHVubmVjZXNzYXJ5IGxvYWQgb24gQUpBWC15CisgICAgICAgIHdlYnNpdGVzLgor CisgICAgICAgIGh0dHBzOi8vYnVncy53ZWJraXQub3JnL3Nob3dfYnVnLmNnaT9pZD00MTgxMwor CisgICAgICAgIFRlc3Q6IGh0dHAvdGVzdHMveG1saHR0cHJlcXVlc3QvY2FjaGUtaGVhZGVycy1h ZnRlci1yZWxvYWQuaHRtbAorCisgICAgICAgICogbG9hZGVyL0ZyYW1lTG9hZGVyLmNwcDoKKyAg ICAgICAgKFdlYkNvcmU6OkZyYW1lTG9hZGVyOjpoYW5kbGVkT25sb2FkRXZlbnRzKTogSWYgdGhl IGxvYWQgdHlwZQorICAgICAgICAgICAgaXMgRnJhbWVMb2FkVHlwZVJlbG9hZCwgY2xlYXIgaXQg dG8gRnJhbWVMb2FkVHlwZVN0YW5kYXJkCisgICAgICAgICAgICBhZnRlciB0aGUgbG9hZCBldmVu dC4KKyAgICAgICAgKiBsb2FkZXIvU3VicmVzb3VyY2VMb2FkZXIuY3BwOgorICAgICAgICAoV2Vi Q29yZTo6U3VicmVzb3VyY2VMb2FkZXI6OmNyZWF0ZSk6IE9ubHkgaW5oZXJpdCBjYWNoZSBwb2xp Y3kKKyAgICAgICAgICAgIGZyb20gdGhlIG9yaWdpbmFsIHJlcXVlc3QgaWYgdGhlIHJlbGV2YW50 IERvY3VtZW50TG9hZGVyCisgICAgICAgICAgICBpcyBzdGlsbCBsb2FkaW5nLgorCiAyMDEwLTA3 LTA3ICBDaHJpcyBGbGVpemFjaCAgPGNmbGVpemFjaEBhcHBsZS5jb20+CiAKICAgICAgICAgUmV2 aWV3ZWQgYnkgQmV0aCBEYWtpbi4KSW5kZXg6IFdlYkNvcmUvbG9hZGVyL0ZyYW1lTG9hZGVyLmNw cAo9PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09Ci0tLSBXZWJDb3JlL2xvYWRlci9GcmFtZUxvYWRlci5jcHAJKHJldmlzaW9u IDYyNzE0KQorKysgV2ViQ29yZS9sb2FkZXIvRnJhbWVMb2FkZXIuY3BwCSh3b3JraW5nIGNvcHkp CkBAIC0yNjAxLDYgKzI2MDEsMTAgQEAKIHZvaWQgRnJhbWVMb2FkZXI6OmhhbmRsZWRPbmxvYWRF dmVudHMoKQogewogICAgIG1fY2xpZW50LT5kaXNwYXRjaERpZEhhbmRsZU9ubG9hZEV2ZW50cygp OworCisgICAgaWYgKG1fbG9hZFR5cGUgPT0gRnJhbWVMb2FkVHlwZVJlbG9hZCkKKyAgICAgICAg bV9sb2FkVHlwZSA9IEZyYW1lTG9hZFR5cGVTdGFuZGFyZDsKKwogI2lmIEVOQUJMRShPRkZMSU5F X1dFQl9BUFBMSUNBVElPTlMpCiAgICAgaWYgKGRvY3VtZW50TG9hZGVyKCkpCiAgICAgICAgIGRv Y3VtZW50TG9hZGVyKCktPmFwcGxpY2F0aW9uQ2FjaGVIb3N0KCktPnN0b3BEZWZlcnJpbmdFdmVu dHMoKTsKSW5kZXg6IFdlYkNvcmUvbG9hZGVyL1N1YnJlc291cmNlTG9hZGVyLmNwcAo9PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09Ci0tLSBXZWJDb3JlL2xvYWRlci9TdWJyZXNvdXJjZUxvYWRlci5jcHAJKHJldmlzaW9uIDYy NzE0KQorKysgV2ViQ29yZS9sb2FkZXIvU3VicmVzb3VyY2VMb2FkZXIuY3BwCSh3b3JraW5nIGNv cHkpCkBAIC04NCw3ICs4NCw4IEBACiAgICAgICAgIG5ld1JlcXVlc3Quc2V0SFRUUFJlZmVycmVy KGZsLT5vdXRnb2luZ1JlZmVycmVyKCkpOwogICAgIEZyYW1lTG9hZGVyOjphZGRIVFRQT3JpZ2lu SWZOZWVkZWQobmV3UmVxdWVzdCwgZmwtPm91dGdvaW5nT3JpZ2luKCkpOwogCi0gICAgLy8gVXNl IHRoZSBvcmlnaW5hbCByZXF1ZXN0J3MgY2FjaGUgcG9saWN5IGZvciB0d28gcmVhc29uczoKKyAg ICAvLyBJZiBpbmhlcml0aW5nIGNhY2hlIHBvbGljeSBmcm9tIGEgRG9jdW1lbnRMb2FkZXIsIHVz ZSB0aGUgb3JpZ2luYWwgcmVxdWVzdCdzCisgICAgLy8gY2FjaGUgcG9saWN5IGZvciB0d28gcmVh c29uczoKICAgICAvLyAxLiBGb3IgUE9TVCByZXF1ZXN0cywgd2UgbXV0YXRlIHRoZSBjYWNoZSBw b2xpY3kgZm9yIHRoZSBtYWluIHJlc291cmNlLAogICAgIC8vICAgIGJ1dCB3ZSBkbyBub3Qgd2Fu dCB0aGlzIHRvIGFwcGx5IHRvIHN1YnJlc291cmNlcwogICAgIC8vIDIuIERlbGVnYXRlcyB0aGF0 IG1vZGlmeSB0aGUgY2FjaGUgcG9saWN5IHVzaW5nIHdpbGxTZW5kUmVxdWVzdDogc2hvdWxkCkBA IC05Miw2ICs5Myw4IEBACiAgICAgLy8gICAgcGVyIHJlcXVlc3QuCiAgICAgaWYgKG5ld1JlcXVl c3QuaXNDb25kaXRpb25hbCgpKQogICAgICAgICBuZXdSZXF1ZXN0LnNldENhY2hlUG9saWN5KFJl bG9hZElnbm9yaW5nQ2FjaGVEYXRhKTsKKyAgICBlbHNlIGlmICghZmwtPmFjdGl2ZURvY3VtZW50 TG9hZGVyKCktPmlzTG9hZGluZ0luQVBJU2Vuc2UoKSkKKyAgICAgICAgbmV3UmVxdWVzdC5zZXRD YWNoZVBvbGljeShVc2VQcm90b2NvbENhY2hlUG9saWN5KTsKICAgICBlbHNlCiAgICAgICAgIG5l d1JlcXVlc3Quc2V0Q2FjaGVQb2xpY3koZmwtPm9yaWdpbmFsUmVxdWVzdCgpLmNhY2hlUG9saWN5 KCkpOwogCkluZGV4OiBMYXlvdXRUZXN0cy9odHRwL3Rlc3RzL3htbGh0dHByZXF1ZXN0L3Jlc291 cmNlcy9wcmludC1jYWNoZS1jb250cm9sLWhlYWRlci5jZ2kKPT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PQotLS0gTGF5b3V0 VGVzdHMvaHR0cC90ZXN0cy94bWxodHRwcmVxdWVzdC9yZXNvdXJjZXMvcHJpbnQtY2FjaGUtY29u dHJvbC1oZWFkZXIuY2dpCShyZXZpc2lvbiAwKQorKysgTGF5b3V0VGVzdHMvaHR0cC90ZXN0cy94 bWxodHRwcmVxdWVzdC9yZXNvdXJjZXMvcHJpbnQtY2FjaGUtY29udHJvbC1oZWFkZXIuY2dpCShy ZXZpc2lvbiAwKQpAQCAtMCwwICsxLDEwIEBACisjIS91c3IvYmluL3BlcmwgLXdUCit1c2Ugc3Ry aWN0OworCitwcmludCAiQ29udGVudC1UeXBlOiB0ZXh0L3BsYWluXG5cbiI7CisKK2ZvcmVhY2gg KGtleXMgJUVOVikgeworICAgIGlmICgkXyA9fiAiSFRUUF9DQUNIRV9DT05UUk9MIikgeworICAg ICAgICBwcmludCAkXyAuICI6ICIgLiAkRU5WeyRffSAuICJcbiI7CisgICAgfQorfQoKUHJvcGVy dHkgY2hhbmdlcyBvbjogTGF5b3V0VGVzdHMvaHR0cC90ZXN0cy94bWxodHRwcmVxdWVzdC9yZXNv dXJjZXMvcHJpbnQtY2FjaGUtY29udHJvbC1oZWFkZXIuY2dpCl9fX19fX19fX19fX19fX19fX19f X19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX18KQWRkZWQ6IHN2 bjpleGVjdXRhYmxlCiAgICsgKgoKSW5kZXg6IExheW91dFRlc3RzL2h0dHAvdGVzdHMveG1saHR0 cHJlcXVlc3QvY2FjaGUtaGVhZGVycy1hZnRlci1yZWxvYWQtZXhwZWN0ZWQudHh0Cj09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT0KLS0tIExheW91dFRlc3RzL2h0dHAvdGVzdHMveG1saHR0cHJlcXVlc3QvY2FjaGUtaGVhZGVy cy1hZnRlci1yZWxvYWQtZXhwZWN0ZWQudHh0CShyZXZpc2lvbiAwKQorKysgTGF5b3V0VGVzdHMv aHR0cC90ZXN0cy94bWxodHRwcmVxdWVzdC9jYWNoZS1oZWFkZXJzLWFmdGVyLXJlbG9hZC1leHBl Y3RlZC50eHQJKHJldmlzaW9uIDApCkBAIC0wLDAgKzEsOCBAQAorVGhpcyB0ZXN0IGRvZXMgdGhl IGZvbGxvd2luZzoKKzEuIExvYWQgYSBwYWdlLCB0cmlnZ2VyIGFuIFhIUi4KKzIuIFJlbG9hZCB0 aGUgcGFnZS4KKzMuIER1bXAgdGhlIGh0dHAgcmVhZGVycyBmb3IgdGhlIHJlbG9hZCBvZiB0aGUg WEhSLgorCitUaGUgWEhSIGlzIHRyaWdnZXIgYWZ0ZXIgb25sb2FkIGhhcyBmaXJlZC4gVGhlcmUg c2hvdWxkIG5vdCBiZSBhbiBIVFRQX0NBQ0hFX0NPTlRST0wgaGVhZGVyIGZvcmNpbmcgcmVsb2Fk LCBzbyBhbiBodHRwIGNhY2hlIGNvdWxkIGxvYWQgdGhlIHJlc291cmNlIHdpdGhvdXQgY29udGFj dGluZyB0aGUgc2VydmVyLgorVGhlIHRlc3QgcGFzc2VzIGlmIHRoZXJlIGlzIG5vIEhUVFBfQ0FD SEVfQ09OVFJPTCBoZWFkZXIgZGlzcGxheWVkIGJlbG93LgorCkluZGV4OiBMYXlvdXRUZXN0cy9o dHRwL3Rlc3RzL3htbGh0dHByZXF1ZXN0L2NhY2hlLWhlYWRlcnMtYWZ0ZXItcmVsb2FkLmh0bWwK PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PQotLS0gTGF5b3V0VGVzdHMvaHR0cC90ZXN0cy94bWxodHRwcmVxdWVzdC9jYWNo ZS1oZWFkZXJzLWFmdGVyLXJlbG9hZC5odG1sCShyZXZpc2lvbiAwKQorKysgTGF5b3V0VGVzdHMv aHR0cC90ZXN0cy94bWxodHRwcmVxdWVzdC9jYWNoZS1oZWFkZXJzLWFmdGVyLXJlbG9hZC5odG1s CShyZXZpc2lvbiAwKQpAQCAtMCwwICsxLDQ0IEBACis8aHRtbD4KKzxib2R5IG9ubG9hZD0ic2V0 VGltZW91dCh0ZXN0LCAxMDApOyI+CitUaGlzIHRlc3QgZG9lcyB0aGUgZm9sbG93aW5nOjxicj4K KzEuIExvYWQgYSBwYWdlLCB0cmlnZ2VyIGFuIFhIUi48YnI+CisyLiBSZWxvYWQgdGhlIHBhZ2Uu PGJyPgorMy4gRHVtcCB0aGUgaHR0cCByZWFkZXJzIGZvciB0aGUgcmVsb2FkIG9mIHRoZSBYSFIu PGJyPjxicj4KK1RoZSBYSFIgaXMgdHJpZ2dlciBhZnRlciBvbmxvYWQgaGFzIGZpcmVkLiBUaGVy ZSBzaG91bGQgbm90IGJlIGFuIEhUVFBfQ0FDSEVfQ09OVFJPTCBoZWFkZXIgZm9yY2luZyByZWxv YWQsIHNvIGFuIGh0dHAgY2FjaGUKK2NvdWxkIGxvYWQgdGhlIHJlc291cmNlIHdpdGhvdXQgY29u dGFjdGluZyB0aGUgc2VydmVyLjxicj4KK1RoZSB0ZXN0IHBhc3NlcyBpZiB0aGVyZSBpcyBubyBI VFRQX0NBQ0hFX0NPTlRST0wgaGVhZGVyIGRpc3BsYXllZCBiZWxvdy48YnI+Cis8ZGl2IGlkPSJj b25zb2xlIj48L2Rpdj4KKzxzY3JpcHQ+CisgICAgaWYgKHdpbmRvdy5sYXlvdXRUZXN0Q29udHJv bGxlcikgeworICAgICAgICBsYXlvdXRUZXN0Q29udHJvbGxlci5kdW1wQXNUZXh0KCk7CisgICAg ICAgIGxheW91dFRlc3RDb250cm9sbGVyLndhaXRVbnRpbERvbmUoKTsKKyAgICB9CisKKyAgICBm dW5jdGlvbiBsb2cobWVzc2FnZSkgeworICAgICAgICBkb2N1bWVudC5nZXRFbGVtZW50QnlJZCgi Y29uc29sZSIpLmFwcGVuZENoaWxkKGRvY3VtZW50LmNyZWF0ZVRleHROb2RlKG1lc3NhZ2UpKTsK KyAgICB9CisKKyAgICB2YXIgeGhyOyAgICAKKyAgICBmdW5jdGlvbiB0ZXN0KCkgeworICAgICAg ICB4aHIgPSBuZXcgWE1MSHR0cFJlcXVlc3QoKTsKKyAgICAgICAgaWYgKGxvY2FsU3RvcmFnZS5y ZWxvYWRlZCkKKyAgICAgICAgICB4aHIub25sb2FkID0gZmluaXNoOworICAgICAgICBlbHNlICAg ICAgICAgCisgICAgICAgICAgeGhyLm9ubG9hZCA9IHJlbG9hZDsKKyAgICAgICAgeGhyLm9wZW4o IkdFVCIsICJyZXNvdXJjZXMvcHJpbnQtY2FjaGUtY29udHJvbC1oZWFkZXIuY2dpIiwgdHJ1ZSk7 CisgICAgICAgIHhoci5zZW5kKG51bGwpOworICAgIH0KKyAgICAKKyAgICBmdW5jdGlvbiByZWxv YWQoKSB7CisgICAgICAgIGxvY2FsU3RvcmFnZS5yZWxvYWRlZCA9IHRydWU7CisgICAgICAgIGxv Y2F0aW9uLnJlbG9hZCh0cnVlKTsKKyAgICB9CisgICAgCisgICAgZnVuY3Rpb24gZmluaXNoKCkg eworICAgICAgICBsb2coeGhyLnJlc3BvbnNlVGV4dCk7CisgICAgICAgIGlmICh3aW5kb3cubGF5 b3V0VGVzdENvbnRyb2xsZXIpCisgICAgICAgICAgICBsYXlvdXRUZXN0Q29udHJvbGxlci5ub3Rp ZnlEb25lKCk7CisgICAgfQorPC9zY3JpcHQ+Cis8L2JvZHk+Cis8L2h0bWw+CkluZGV4OiBMYXlv dXRUZXN0cy9DaGFuZ2VMb2cKPT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PQotLS0gTGF5b3V0VGVzdHMvQ2hhbmdlTG9nCShy ZXZpc2lvbiA2MjcyOSkKKysrIExheW91dFRlc3RzL0NoYW5nZUxvZwkod29ya2luZyBjb3B5KQpA QCAtMSwzICsxLDE2IEBACisyMDEwLTA3LTA3ICBOYXRlIENoYXBpbiAgPGphcGhldEBjaHJvbWl1 bS5vcmc+CisKKyAgICAgICAgUmV2aWV3ZWQgYnkgTk9CT0RZIChPT1BTISkuCisKKyAgICAgICAg VGVzdCB0aGF0IENhY2hlLUNvbnRyb2wgaGVhZGVycyBhcmUgbm90IGluY2x1ZGVkIG9uIFhIUnMg dGhhdCBvY2N1cgorICAgICAgICBhZnRlciB0aGUgbG9hZCBldmVudC4KKworICAgICAgICBodHRw czovL2J1Z3Mud2Via2l0Lm9yZy9zaG93X2J1Zy5jZ2k/aWQ9NDE4MTMKKworICAgICAgICAqIGh0 dHAvdGVzdHMveG1saHR0cHJlcXVlc3QvY2FjaGUtaGVhZGVycy1hZnRlci1yZWxvYWQtZXhwZWN0 ZWQudHh0OiBBZGRlZC4KKyAgICAgICAgKiBodHRwL3Rlc3RzL3htbGh0dHByZXF1ZXN0L2NhY2hl LWhlYWRlcnMtYWZ0ZXItcmVsb2FkLmh0bWw6IEFkZGVkLgorICAgICAgICAqIGh0dHAvdGVzdHMv eG1saHR0cHJlcXVlc3QvcmVzb3VyY2VzL3ByaW50LWNhY2hlLWNvbnRyb2wtaGVhZGVyLmNnaTog QWRkZWQuCisKIDIwMTAtMDctMDcgIFBldGVyIEthc3RpbmcgIDxwa2FzdGluZ0Bnb29nbGUuY29t PgogCiAgICAgICAgIE5vdCByZXZpZXdlZC4gIENocm9taXVtIHRlc3QgZXhwZWN0YXRpb24gdXBk YXRlLgo= 61100 2010-07-09 14:41:34 -0700 2010-07-09 17:08:01 -0700 patch #2 xhr-cache2.txt text/plain 9790 japhet SW5kZXg6IFdlYkNvcmUvQ2hhbmdlTG9nCj09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT0KLS0tIFdlYkNvcmUvQ2hhbmdlTG9n CShyZXZpc2lvbiA2Mjk4MikKKysrIFdlYkNvcmUvQ2hhbmdlTG9nCSh3b3JraW5nIGNvcHkpCkBA IC0xLDMgKzEsMjQgQEAKKzIwMTAtMDctMDkgIE5hdGUgQ2hhcGluICA8amFwaGV0QGNocm9taXVt Lm9yZz4KKworICAgICAgICBSZXZpZXdlZCBieSBOT0JPRFkgKE9PUFMhKS4KKworICAgICAgICBF bnN1cmUgdGhhdCBhIGNhY2hlIHBvbGljeSB0aGF0IGZvcmNlcyB2YWxpZGF0aW9uIGlzIGNsZWFy ZWQgb25jZQorICAgICAgICB0aGUgbG9hZCBldmVudCBpcyBmaXJlZCwgcmF0aGVyIHRoYW4gb25s eSBkb2luZyBzbyBhdCB0aGUgbmV4dAorICAgICAgICBuYXZpZ2F0aW9uLiBUaGlzIGxlYWRzIHRv IGEgbG90IG9mIHVubmVjZXNzYXJ5IGxvYWQgb24gQUpBWC15CisgICAgICAgIHdlYnNpdGVzLgor CisgICAgICAgIGh0dHBzOi8vYnVncy53ZWJraXQub3JnL3Nob3dfYnVnLmNnaT9pZD00MTgxMwor CisgICAgICAgIFRlc3Q6IGh0dHAvdGVzdHMveG1saHR0cHJlcXVlc3QvY2FjaGUtaGVhZGVycy1h ZnRlci1yZWxvYWQuaHRtbAorCisgICAgICAgICogbG9hZGVyL0ZyYW1lTG9hZGVyLmNwcDoKKyAg ICAgICAgKFdlYkNvcmU6OkZyYW1lTG9hZGVyOjpoYW5kbGVkT25sb2FkRXZlbnRzKTogUmVzZXQg bV9sb2FkVHlwZSB0byBGcmFtZUxvYWRUeXBlU3RhbmRhcmQuCisgICAgICAgIChXZWJDb3JlOjpG cmFtZUxvYWRlcjo6YWRkRXh0cmFGaWVsZHNUb1JlcXVlc3QpOiBPbmx5IHJlc3BlY3QgdGhlIG9y aWdpbmFsIHJlcXVlc3QncyBjYWNoZSBwb2xpY3kgaWYgdGhlCisgICAgICAgICAgICBEb2N1bWVu dExvYWRlciBpcyBzdGlsbCBsb2FkaW5nLCBhbmQgaGFuZGxlIHRoZSBvdGhlciBjYWNoZSBwb2xp Y3kgc2V0dGluZ3MgdGhhdCB3ZXJlIHNjYXR0ZXJlZCBhcm91bmQgdGhlIGxvYWRlci4KKyAgICAg ICAgKFdlYkNvcmU6OkZyYW1lTG9hZGVyOjpsb2FkUmVzb3VyY2VTeW5jaHJvbm91c2x5KTogTWVy Z2UgY2FjaGVQb2xpY3kgc2V0dGluZyBpbnRvIEZyYW1lTG9hZGVyOjphZGRFeHRyYUZpZWxkc1Rv UmVxdWVzdC4KKyAgICAgICAgKiBsb2FkZXIvU3VicmVzb3VyY2VMb2FkZXIuY3BwOgorICAgICAg ICAoV2ViQ29yZTo6U3VicmVzb3VyY2VMb2FkZXI6OmNyZWF0ZSk6IE1lcmdlIGNhY2hlUG9saWN5 IHNldHRpbmcgaW50byBGcmFtZUxvYWRlcjo6YWRkRXh0cmFGaWVsZHNUb1JlcXVlc3QuCisKIDIw MTAtMDctMDkgIER1bWl0cnUgRGFuaWxpdWMgIDxkdW1pQGNocm9taXVtLm9yZz4KIAogICAgICAg ICBVbnJldmlld2VkLCBidXQgcHJlLWFwcHJvdmVkIGJ5IEVyaWMgU2VpZGVsLgpJbmRleDogV2Vi Q29yZS9sb2FkZXIvRnJhbWVMb2FkZXIuY3BwCj09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT0KLS0tIFdlYkNvcmUvbG9hZGVy L0ZyYW1lTG9hZGVyLmNwcAkocmV2aXNpb24gNjI4NjQpCisrKyBXZWJDb3JlL2xvYWRlci9GcmFt ZUxvYWRlci5jcHAJKHdvcmtpbmcgY29weSkKQEAgLTI2MDEsNiArMjYwMSw5IEBACiB2b2lkIEZy YW1lTG9hZGVyOjpoYW5kbGVkT25sb2FkRXZlbnRzKCkKIHsKICAgICBtX2NsaWVudC0+ZGlzcGF0 Y2hEaWRIYW5kbGVPbmxvYWRFdmVudHMoKTsKKworICAgIG1fbG9hZFR5cGUgPSBGcmFtZUxvYWRU eXBlU3RhbmRhcmQ7CisKICNpZiBFTkFCTEUoT0ZGTElORV9XRUJfQVBQTElDQVRJT05TKQogICAg IGlmIChkb2N1bWVudExvYWRlcigpKQogICAgICAgICBkb2N1bWVudExvYWRlcigpLT5hcHBsaWNh dGlvbkNhY2hlSG9zdCgpLT5zdG9wRGVmZXJyaW5nRXZlbnRzKCk7CkBAIC0yNjczLDYgKzI2NzYs MTMgQEAKIAogICAgIGFwcGx5VXNlckFnZW50KHJlcXVlc3QpOwogICAgIAorICAgIC8vIElmIHdl IGluaGVyaXQgY2FjaGUgcG9saWN5IGZyb20gYSBtYWluIHJlc291cmNlLCB3ZSB1c2UgdGhlIERv Y3VtZW50TG9hZGVyJ3MgCisgICAgLy8gb3JpZ2luYWwgcmVxdWVzdCBjYWNoZSBwb2xpY3kgZm9y IHR3byByZWFzb25zOgorICAgIC8vIDEuIEZvciBQT1NUIHJlcXVlc3RzLCB3ZSBtdXRhdGUgdGhl IGNhY2hlIHBvbGljeSBmb3IgdGhlIG1haW4gcmVzb3VyY2UsCisgICAgLy8gICAgYnV0IHdlIGRv IG5vdCB3YW50IHRoaXMgdG8gYXBwbHkgdG8gc3VicmVzb3VyY2VzCisgICAgLy8gMi4gRGVsZWdh dGVzIHRoYXQgbW9kaWZ5IHRoZSBjYWNoZSBwb2xpY3kgdXNpbmcgd2lsbFNlbmRSZXF1ZXN0OiBz aG91bGQKKyAgICAvLyAgICBub3QgYWZmZWN0IGFueSBvdGhlciByZXNvdXJjZXMuIFN1Y2ggY2hh bmdlcyBuZWVkIHRvIGJlIGRvbmUKKyAgICAvLyAgICBwZXIgcmVxdWVzdC4KICAgICBpZiAobG9h ZFR5cGUgPT0gRnJhbWVMb2FkVHlwZVJlbG9hZCkgewogICAgICAgICByZXF1ZXN0LnNldENhY2hl UG9saWN5KFJlbG9hZElnbm9yaW5nQ2FjaGVEYXRhKTsKICAgICAgICAgcmVxdWVzdC5zZXRIVFRQ SGVhZGVyRmllbGQoIkNhY2hlLUNvbnRyb2wiLCAibWF4LWFnZT0wIik7CkBAIC0yNjgwLDggKzI2 OTAsMTIgQEAKICAgICAgICAgcmVxdWVzdC5zZXRDYWNoZVBvbGljeShSZWxvYWRJZ25vcmluZ0Nh Y2hlRGF0YSk7CiAgICAgICAgIHJlcXVlc3Quc2V0SFRUUEhlYWRlckZpZWxkKCJDYWNoZS1Db250 cm9sIiwgIm5vLWNhY2hlIik7CiAgICAgICAgIHJlcXVlc3Quc2V0SFRUUEhlYWRlckZpZWxkKCJQ cmFnbWEiLCAibm8tY2FjaGUiKTsKLSAgICB9IGVsc2UgaWYgKGlzQmFja0ZvcndhcmRMb2FkVHlw ZShsb2FkVHlwZSkgJiYgIXJlcXVlc3QudXJsKCkucHJvdG9jb2xJcygiaHR0cHMiKSkKKyAgICB9 IGVsc2UgaWYgKHJlcXVlc3QuaXNDb25kaXRpb25hbCgpKQorICAgICAgICByZXF1ZXN0LnNldENh Y2hlUG9saWN5KFJlbG9hZElnbm9yaW5nQ2FjaGVEYXRhKTsKKyAgICBlbHNlIGlmIChpc0JhY2tG b3J3YXJkTG9hZFR5cGUobG9hZFR5cGUpICYmICFyZXF1ZXN0LnVybCgpLnByb3RvY29sSXMoImh0 dHBzIikpCiAgICAgICAgIHJlcXVlc3Quc2V0Q2FjaGVQb2xpY3koUmV0dXJuQ2FjaGVEYXRhRWxz ZUxvYWQpOworICAgIGVsc2UgaWYgKCFtYWluUmVzb3VyY2UgJiYgZG9jdW1lbnRMb2FkZXIoKS0+ aXNMb2FkaW5nSW5BUElTZW5zZSgpKQorICAgICAgICByZXF1ZXN0LnNldENhY2hlUG9saWN5KGRv Y3VtZW50TG9hZGVyKCktPm9yaWdpbmFsUmVxdWVzdCgpLmNhY2hlUG9saWN5KCkpOwogICAgIAog ICAgIGlmIChtYWluUmVzb3VyY2UpCiAgICAgICAgIHJlcXVlc3Quc2V0SFRUUEFjY2VwdChkZWZh dWx0QWNjZXB0SGVhZGVyKTsKQEAgLTI3NzgsMTcgKzI3OTIsNiBAQAogICAgIFJlc291cmNlUmVx dWVzdCBpbml0aWFsUmVxdWVzdCA9IHJlcXVlc3Q7CiAgICAgaW5pdGlhbFJlcXVlc3Quc2V0VGlt ZW91dEludGVydmFsKDEwKTsKICAgICAKLSAgICAvLyBVc2UgdGhlIG9yaWdpbmFsIHJlcXVlc3Qn cyBjYWNoZSBwb2xpY3kgZm9yIHR3byByZWFzb25zOgotICAgIC8vIDEuIEZvciBQT1NUIHJlcXVl c3RzLCB3ZSBtdXRhdGUgdGhlIGNhY2hlIHBvbGljeSBmb3IgdGhlIG1haW4gcmVzb3VyY2UsCi0g ICAgLy8gICAgYnV0IHdlIGRvIG5vdCB3YW50IHRoaXMgdG8gYXBwbHkgdG8gc3VicmVzb3VyY2Vz Ci0gICAgLy8gMi4gRGVsZWdhdGVzIHRoYXQgbW9kaWZ5IHRoZSBjYWNoZSBwb2xpY3kgdXNpbmcg d2lsbFNlbmRSZXF1ZXN0OiBzaG91bGQKLSAgICAvLyAgICBub3QgYWZmZWN0IGFueSBvdGhlciBy ZXNvdXJjZXMuIFN1Y2ggY2hhbmdlcyBuZWVkIHRvIGJlIGRvbmUKLSAgICAvLyAgICBwZXIgcmVx dWVzdC4KLSAgICBpZiAoaW5pdGlhbFJlcXVlc3QuaXNDb25kaXRpb25hbCgpKQotICAgICAgICBp bml0aWFsUmVxdWVzdC5zZXRDYWNoZVBvbGljeShSZWxvYWRJZ25vcmluZ0NhY2hlRGF0YSk7Ci0g ICAgZWxzZQotICAgICAgICBpbml0aWFsUmVxdWVzdC5zZXRDYWNoZVBvbGljeShvcmlnaW5hbFJl cXVlc3QoKS5jYWNoZVBvbGljeSgpKTsKLSAgICAKICAgICBpZiAoIXJlZmVycmVyLmlzRW1wdHko KSkKICAgICAgICAgaW5pdGlhbFJlcXVlc3Quc2V0SFRUUFJlZmVycmVyKHJlZmVycmVyKTsKICAg ICBhZGRIVFRQT3JpZ2luSWZOZWVkZWQoaW5pdGlhbFJlcXVlc3QsIG91dGdvaW5nT3JpZ2luKCkp OwpAQCAtMjc5Niw2ICsyNzk5LDggQEAKICAgICBpZiAoUGFnZSogcGFnZSA9IG1fZnJhbWUtPnBh Z2UoKSkKICAgICAgICAgaW5pdGlhbFJlcXVlc3Quc2V0Rmlyc3RQYXJ0eUZvckNvb2tpZXMocGFn ZS0+bWFpbkZyYW1lKCktPmxvYWRlcigpLT5kb2N1bWVudExvYWRlcigpLT5yZXF1ZXN0KCkudXJs KCkpOwogICAgIGluaXRpYWxSZXF1ZXN0LnNldEhUVFBVc2VyQWdlbnQoY2xpZW50KCktPnVzZXJB Z2VudChyZXF1ZXN0LnVybCgpKSk7CisgICAgCisgICAgYWRkRXh0cmFGaWVsZHNUb1N1YnJlc291 cmNlUmVxdWVzdChpbml0aWFsUmVxdWVzdCk7CiAKICAgICB1bnNpZ25lZCBsb25nIGlkZW50aWZp ZXIgPSAwOyAgICAKICAgICBSZXNvdXJjZVJlcXVlc3QgbmV3UmVxdWVzdChpbml0aWFsUmVxdWVz dCk7CkluZGV4OiBXZWJDb3JlL2xvYWRlci9TdWJyZXNvdXJjZUxvYWRlci5jcHAKPT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PQotLS0gV2ViQ29yZS9sb2FkZXIvU3VicmVzb3VyY2VMb2FkZXIuY3BwCShyZXZpc2lvbiA2Mjg2 NCkKKysrIFdlYkNvcmUvbG9hZGVyL1N1YnJlc291cmNlTG9hZGVyLmNwcAkod29ya2luZyBjb3B5 KQpAQCAtODQsMTcgKzg0LDYgQEAKICAgICAgICAgbmV3UmVxdWVzdC5zZXRIVFRQUmVmZXJyZXIo ZmwtPm91dGdvaW5nUmVmZXJyZXIoKSk7CiAgICAgRnJhbWVMb2FkZXI6OmFkZEhUVFBPcmlnaW5J Zk5lZWRlZChuZXdSZXF1ZXN0LCBmbC0+b3V0Z29pbmdPcmlnaW4oKSk7CiAKLSAgICAvLyBVc2Ug dGhlIG9yaWdpbmFsIHJlcXVlc3QncyBjYWNoZSBwb2xpY3kgZm9yIHR3byByZWFzb25zOgotICAg IC8vIDEuIEZvciBQT1NUIHJlcXVlc3RzLCB3ZSBtdXRhdGUgdGhlIGNhY2hlIHBvbGljeSBmb3Ig dGhlIG1haW4gcmVzb3VyY2UsCi0gICAgLy8gICAgYnV0IHdlIGRvIG5vdCB3YW50IHRoaXMgdG8g YXBwbHkgdG8gc3VicmVzb3VyY2VzCi0gICAgLy8gMi4gRGVsZWdhdGVzIHRoYXQgbW9kaWZ5IHRo ZSBjYWNoZSBwb2xpY3kgdXNpbmcgd2lsbFNlbmRSZXF1ZXN0OiBzaG91bGQKLSAgICAvLyAgICBu b3QgYWZmZWN0IGFueSBvdGhlciByZXNvdXJjZXMuIFN1Y2ggY2hhbmdlcyBuZWVkIHRvIGJlIGRv bmUKLSAgICAvLyAgICBwZXIgcmVxdWVzdC4KLSAgICBpZiAobmV3UmVxdWVzdC5pc0NvbmRpdGlv bmFsKCkpCi0gICAgICAgIG5ld1JlcXVlc3Quc2V0Q2FjaGVQb2xpY3koUmVsb2FkSWdub3JpbmdD YWNoZURhdGEpOwotICAgIGVsc2UKLSAgICAgICAgbmV3UmVxdWVzdC5zZXRDYWNoZVBvbGljeShm bC0+b3JpZ2luYWxSZXF1ZXN0KCkuY2FjaGVQb2xpY3koKSk7Ci0KICAgICBmbC0+YWRkRXh0cmFG aWVsZHNUb1N1YnJlc291cmNlUmVxdWVzdChuZXdSZXF1ZXN0KTsKIAogICAgIFJlZlB0cjxTdWJy ZXNvdXJjZUxvYWRlcj4gc3VibG9hZGVyKGFkb3B0UmVmKG5ldyBTdWJyZXNvdXJjZUxvYWRlcihm cmFtZSwgY2xpZW50LCBzZW5kUmVzb3VyY2VMb2FkQ2FsbGJhY2tzLCBzaG91bGRDb250ZW50U25p ZmYpKSk7CkluZGV4OiBMYXlvdXRUZXN0cy9odHRwL3Rlc3RzL3htbGh0dHByZXF1ZXN0L3Jlc291 cmNlcy9wcmludC1jYWNoZS1jb250cm9sLWhlYWRlci5jZ2kKPT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PQotLS0gTGF5b3V0 VGVzdHMvaHR0cC90ZXN0cy94bWxodHRwcmVxdWVzdC9yZXNvdXJjZXMvcHJpbnQtY2FjaGUtY29u dHJvbC1oZWFkZXIuY2dpCShyZXZpc2lvbiAwKQorKysgTGF5b3V0VGVzdHMvaHR0cC90ZXN0cy94 bWxodHRwcmVxdWVzdC9yZXNvdXJjZXMvcHJpbnQtY2FjaGUtY29udHJvbC1oZWFkZXIuY2dpCShy ZXZpc2lvbiAwKQpAQCAtMCwwICsxLDEwIEBACisjIS91c3IvYmluL3BlcmwgLXdUCit1c2Ugc3Ry aWN0OworCitwcmludCAiQ29udGVudC1UeXBlOiB0ZXh0L3BsYWluXG5cbiI7CisKK2ZvcmVhY2gg KGtleXMgJUVOVikgeworICAgIGlmICgkXyA9fiAiSFRUUF9DQUNIRV9DT05UUk9MIikgeworICAg ICAgICBwcmludCAkXyAuICI6ICIgLiAkRU5WeyRffSAuICJcbiI7CisgICAgfQorfQoKUHJvcGVy dHkgY2hhbmdlcyBvbjogTGF5b3V0VGVzdHMvaHR0cC90ZXN0cy94bWxodHRwcmVxdWVzdC9yZXNv dXJjZXMvcHJpbnQtY2FjaGUtY29udHJvbC1oZWFkZXIuY2dpCl9fX19fX19fX19fX19fX19fX19f X19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX18KQWRkZWQ6IHN2 bjpleGVjdXRhYmxlCiAgICsgKgoKSW5kZXg6IExheW91dFRlc3RzL2h0dHAvdGVzdHMveG1saHR0 cHJlcXVlc3QvY2FjaGUtaGVhZGVycy1hZnRlci1yZWxvYWQtZXhwZWN0ZWQudHh0Cj09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT0KLS0tIExheW91dFRlc3RzL2h0dHAvdGVzdHMveG1saHR0cHJlcXVlc3QvY2FjaGUtaGVhZGVy cy1hZnRlci1yZWxvYWQtZXhwZWN0ZWQudHh0CShyZXZpc2lvbiAwKQorKysgTGF5b3V0VGVzdHMv aHR0cC90ZXN0cy94bWxodHRwcmVxdWVzdC9jYWNoZS1oZWFkZXJzLWFmdGVyLXJlbG9hZC1leHBl Y3RlZC50eHQJKHJldmlzaW9uIDApCkBAIC0wLDAgKzEsOCBAQAorVGhpcyB0ZXN0IGRvZXMgdGhl IGZvbGxvd2luZzoKKzEuIExvYWQgYSBwYWdlLCB0cmlnZ2VyIGFuIFhIUi4KKzIuIFJlbG9hZCB0 aGUgcGFnZS4KKzMuIER1bXAgdGhlIGh0dHAgcmVhZGVycyBmb3IgdGhlIHJlbG9hZCBvZiB0aGUg WEhSLgorCitUaGUgWEhSIGlzIHRyaWdnZXIgYWZ0ZXIgb25sb2FkIGhhcyBmaXJlZC4gVGhlcmUg c2hvdWxkIG5vdCBiZSBhbiBIVFRQX0NBQ0hFX0NPTlRST0wgaGVhZGVyIGZvcmNpbmcgcmVsb2Fk LCBzbyBhbiBodHRwIGNhY2hlIGNvdWxkIGxvYWQgdGhlIHJlc291cmNlIHdpdGhvdXQgY29udGFj dGluZyB0aGUgc2VydmVyLgorVGhlIHRlc3QgcGFzc2VzIGlmIHRoZXJlIGlzIG5vIEhUVFBfQ0FD SEVfQ09OVFJPTCBoZWFkZXIgZGlzcGxheWVkIGJlbG93LgorCkluZGV4OiBMYXlvdXRUZXN0cy9o dHRwL3Rlc3RzL3htbGh0dHByZXF1ZXN0L2NhY2hlLWhlYWRlcnMtYWZ0ZXItcmVsb2FkLmh0bWwK PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PQotLS0gTGF5b3V0VGVzdHMvaHR0cC90ZXN0cy94bWxodHRwcmVxdWVzdC9jYWNo ZS1oZWFkZXJzLWFmdGVyLXJlbG9hZC5odG1sCShyZXZpc2lvbiAwKQorKysgTGF5b3V0VGVzdHMv aHR0cC90ZXN0cy94bWxodHRwcmVxdWVzdC9jYWNoZS1oZWFkZXJzLWFmdGVyLXJlbG9hZC5odG1s CShyZXZpc2lvbiAwKQpAQCAtMCwwICsxLDQ0IEBACis8aHRtbD4KKzxib2R5IG9ubG9hZD0ic2V0 VGltZW91dCh0ZXN0LCAxMDApOyI+CitUaGlzIHRlc3QgZG9lcyB0aGUgZm9sbG93aW5nOjxicj4K KzEuIExvYWQgYSBwYWdlLCB0cmlnZ2VyIGFuIFhIUi48YnI+CisyLiBSZWxvYWQgdGhlIHBhZ2Uu PGJyPgorMy4gRHVtcCB0aGUgaHR0cCByZWFkZXJzIGZvciB0aGUgcmVsb2FkIG9mIHRoZSBYSFIu PGJyPjxicj4KK1RoZSBYSFIgaXMgdHJpZ2dlciBhZnRlciBvbmxvYWQgaGFzIGZpcmVkLiBUaGVy ZSBzaG91bGQgbm90IGJlIGFuIEhUVFBfQ0FDSEVfQ09OVFJPTCBoZWFkZXIgZm9yY2luZyByZWxv YWQsIHNvIGFuIGh0dHAgY2FjaGUKK2NvdWxkIGxvYWQgdGhlIHJlc291cmNlIHdpdGhvdXQgY29u dGFjdGluZyB0aGUgc2VydmVyLjxicj4KK1RoZSB0ZXN0IHBhc3NlcyBpZiB0aGVyZSBpcyBubyBI VFRQX0NBQ0hFX0NPTlRST0wgaGVhZGVyIGRpc3BsYXllZCBiZWxvdy48YnI+Cis8ZGl2IGlkPSJj b25zb2xlIj48L2Rpdj4KKzxzY3JpcHQ+CisgICAgaWYgKHdpbmRvdy5sYXlvdXRUZXN0Q29udHJv bGxlcikgeworICAgICAgICBsYXlvdXRUZXN0Q29udHJvbGxlci5kdW1wQXNUZXh0KCk7CisgICAg ICAgIGxheW91dFRlc3RDb250cm9sbGVyLndhaXRVbnRpbERvbmUoKTsKKyAgICB9CisKKyAgICBm dW5jdGlvbiBsb2cobWVzc2FnZSkgeworICAgICAgICBkb2N1bWVudC5nZXRFbGVtZW50QnlJZCgi Y29uc29sZSIpLmFwcGVuZENoaWxkKGRvY3VtZW50LmNyZWF0ZVRleHROb2RlKG1lc3NhZ2UpKTsK KyAgICB9CisKKyAgICB2YXIgeGhyOyAgICAKKyAgICBmdW5jdGlvbiB0ZXN0KCkgeworICAgICAg ICB4aHIgPSBuZXcgWE1MSHR0cFJlcXVlc3QoKTsKKyAgICAgICAgaWYgKGxvY2FsU3RvcmFnZS5y ZWxvYWRlZCkKKyAgICAgICAgICB4aHIub25sb2FkID0gZmluaXNoOworICAgICAgICBlbHNlICAg ICAgICAgCisgICAgICAgICAgeGhyLm9ubG9hZCA9IHJlbG9hZDsKKyAgICAgICAgeGhyLm9wZW4o IkdFVCIsICJyZXNvdXJjZXMvcHJpbnQtY2FjaGUtY29udHJvbC1oZWFkZXIuY2dpIiwgdHJ1ZSk7 CisgICAgICAgIHhoci5zZW5kKG51bGwpOworICAgIH0KKyAgICAKKyAgICBmdW5jdGlvbiByZWxv YWQoKSB7CisgICAgICAgIGxvY2FsU3RvcmFnZS5yZWxvYWRlZCA9IHRydWU7CisgICAgICAgIGxv Y2F0aW9uLnJlbG9hZCh0cnVlKTsKKyAgICB9CisgICAgCisgICAgZnVuY3Rpb24gZmluaXNoKCkg eworICAgICAgICBsb2coeGhyLnJlc3BvbnNlVGV4dCk7CisgICAgICAgIGlmICh3aW5kb3cubGF5 b3V0VGVzdENvbnRyb2xsZXIpCisgICAgICAgICAgICBsYXlvdXRUZXN0Q29udHJvbGxlci5ub3Rp ZnlEb25lKCk7CisgICAgfQorPC9zY3JpcHQ+Cis8L2JvZHk+Cis8L2h0bWw+CkluZGV4OiBMYXlv dXRUZXN0cy9DaGFuZ2VMb2cKPT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PQotLS0gTGF5b3V0VGVzdHMvQ2hhbmdlTG9nCShy ZXZpc2lvbiA2Mjg2NCkKKysrIExheW91dFRlc3RzL0NoYW5nZUxvZwkod29ya2luZyBjb3B5KQpA QCAtMSwzICsxLDE2IEBACisyMDEwLTA3LTA3ICBOYXRlIENoYXBpbiAgPGphcGhldEBjaHJvbWl1 bS5vcmc+CisKKyAgICAgICAgUmV2aWV3ZWQgYnkgTk9CT0RZIChPT1BTISkuCisKKyAgICAgICAg VGVzdCB0aGF0IENhY2hlLUNvbnRyb2wgaGVhZGVycyBhcmUgbm90IGluY2x1ZGVkIG9uIFhIUnMg dGhhdCBvY2N1cgorICAgICAgICBhZnRlciB0aGUgbG9hZCBldmVudC4KKworICAgICAgICBodHRw czovL2J1Z3Mud2Via2l0Lm9yZy9zaG93X2J1Zy5jZ2k/aWQ9NDE4MTMKKworICAgICAgICAqIGh0 dHAvdGVzdHMveG1saHR0cHJlcXVlc3QvY2FjaGUtaGVhZGVycy1hZnRlci1yZWxvYWQtZXhwZWN0 ZWQudHh0OiBBZGRlZC4KKyAgICAgICAgKiBodHRwL3Rlc3RzL3htbGh0dHByZXF1ZXN0L2NhY2hl LWhlYWRlcnMtYWZ0ZXItcmVsb2FkLmh0bWw6IEFkZGVkLgorICAgICAgICAqIGh0dHAvdGVzdHMv eG1saHR0cHJlcXVlc3QvcmVzb3VyY2VzL3ByaW50LWNhY2hlLWNvbnRyb2wtaGVhZGVyLmNnaTog QWRkZWQuCisKIDIwMTAtMDctMDggIEdhdmluIEJhcnJhY2xvdWdoICA8YmFycmFjbG91Z2hAYXBw bGUuY29tPgogCiAgICAgICAgIFJldmlld2VkIGJ5IFNhbSBXZWluaWcuCg==