29021 2009-09-07 22:05:31 -0700 Explore allowing pages with unload handlers into the Page Cache 2010-01-26 07:51:15 -0800 1 1 1 Unclassified WebKit Page Loading 528+ (Nightly build) All All RESOLVED FIXED InRadar P2 Normal --- 1 beidson beidson ap ddkilzer emacemac7 eric jaffathecake klobag koivisto laszlo.gombos matafagafo mrowe paul sam timothy tonikitoo oldest_to_newest 145261 0 beidson 2009-09-07 22:05:31 -0700 We should explore allowing pages with unload handlers into the Page Cache. Historically, we decided to discount these pages under the assumption that unload handlers do critical work. Firefox seemed to follow suite a few years later, as evidenced by https://developer.mozilla.org/En/Using_Firefox_1.5_caching In recent months, while exploring enhancements we can make to the page cache, it's become clear that *many* pages are discounted from our page cache because of unload handlers. And - quite frankly - most of them don't do important work, don't do interesting work, or don't even do any work at all. We've now added the pagehide/pageshow events that Firefox added to help developers work around this page caching limitation. Also, the landscape of the web has changed quite a bit since 2003. We should explore whether or not keeping this limitation is worthwhile. This exploration might be as simple as checking in the one line change that will allow these pages in to the page cache, and watch and see if anything important breaks. In radar as <rdar://problem/7196485> 146045 1 sam 2009-09-09 18:18:18 -0700 I would prefer if this went behind an ENABLE #define to allow ports to easily choose their behavior while this is still under investigation. In general, this seems like it could impose a functional regression and one with subtle and hard to diagnose fallout at that, and thus giving flexibility during the exploration seems prudent. 147248 2 39585 beidson 2009-09-14 20:05:55 -0700 Created attachment 39585 Disable the unload handler check on Mac + Windows 147249 3 beidson 2009-09-14 20:06:56 -0700 Once this gets checked in, I have a blog post prepared that will discuss the state of the page cache, and put a call out to site/library developers to experiment and comment. 147250 4 39586 beidson 2009-09-14 20:14:18 -0700 Created attachment 39586 Slightly better preprocessor style this time 147251 5 beidson 2009-09-14 21:33:52 -0700 Committed r48380: <http://trac.webkit.org/changeset/48380> 147258 6 koivisto 2009-09-14 23:15:27 -0700 I think you mean http://trac.webkit.org/changeset/48388 147259 7 beidson 2009-09-14 23:27:15 -0700 Yes, I did, and I'm never using bugzilla-tool again! 147308 8 eric 2009-09-15 09:40:22 -0700 (In reply to comment #7) > Yes, I did, and I'm never using bugzilla-tool again! Hum. I've never seen it post the wrong revision number like that, but I would love to find out why! I'll follow up with you via email and file/fix any necessary bugs. 147440 9 beidson 2009-09-15 18:07:56 -0700 Re-opening, as this bug needs to track the possibility of putting this check back in to place. 147740 10 39681 beidson 2009-09-16 23:17:15 -0700 Created attachment 39681 Don't fire unload events when leaving a page if the document has been cached. As long as we're running this experiment, lets run it right! Don't fire unload events if the document was put in the Page Cache. 147742 11 39681 sam 2009-09-16 23:29:23 -0700 Comment on attachment 39681 Don't fire unload events when leaving a page if the document has been cached. You sure this can't be tested? 147743 12 beidson 2009-09-16 23:35:54 -0700 I can't think of a good way right now that wouldn't some *very* ugly hacking in WebCore simply for DRTs benefit. Perhaps it's my lack of caffiene and sleep. If I think of one in the morning, I'll add it. 147744 13 beidson 2009-09-16 23:37:03 -0700 http://trac.webkit.org/changeset/48456 147958 14 jaffathecake 2009-09-17 15:39:37 -0700 You could save the page to the page cache, then fire the unload event and navigate away. If back is clicked, the page will be returned in the state it was BEFORE unload was fired. This gives you the best of both worlds. (this was suggested in Chrome's bug tracker) If you're going to fire unload when the page is removed from memory (rather than when it is navigated away from) you need to be careful with "Your work is unsaved, are you sure you want to navigate away?" type messages. Applications like Gmail use these. However, the solution above works around this. Also, (and webkit may do this already) you need to make sure the element that the mouse is currently over has its mouseout event fired. Else you might get sticky hover effects. It's great that this is being looked into, as many js libraries use the unload event to clean up in IE & prevent the sticky hover issue, breaking the page cache. Jake. 148965 15 paul 2009-09-21 23:02:49 -0700 Hey, I just read the blog post over on Sufin' Safari (really good btw) and I thought I'd throw out our use case for the onunload event as a data point. I work at Meebo and we have two pieces of JavaScript that make heavy use of this event, meebo.com and a Meebo chat bar that is syndicated on partner Web sites. In both of these cases we're using long polling to maintain a connection to our servers to power IM presence for the user. We use the onunload event to send a "quit" command to the server to log out of IM when the page is navigated away. For meebo.com (which is a destination IM Web site) we also use the onbeforeunload event to warn the user that navigating from the page will cause their session to be disconnected. When the user navigates away from the page we rely on the onunload event in order to disconnect the user from their session, and I actually think in this case it would be undesirable for the quick "Page Cache" backwards navigation, as this would imply that we would reconnect you to IM, when I feel like navigating away from this page should probably tear down the page and land you back on a blank slate if you were to navigate back. Not firing the onunload event is also undesirable as you'll appear to stay online for another minute or two after you've left before our server times out the session. For the Meebo chat bar (which is conceptually very similar to Facebook's chat bar) your presence should actually follow you based on whether you're on the page or not so in this case implementing the pageshow/pagehide events rather than the onunload event would be ideal, and I hope that we'll be able to do this in the not too distant future! I hope these use cases are useful! 149310 16 timothy 2009-09-22 21:01:56 -0700 On the flipside, meebo.com could use pagehide to disconnect and pageshow to reconnect you. This would let you keep the same IM "windows" open, and all other sate. That is what I would expect as a meebo.com user. But I guess there are security concerns… 152366 17 eric 2009-10-05 11:02:47 -0700 I assume this can be closed now that it's landed? 152394 18 beidson 2009-10-05 11:30:10 -0700 No. Unfortunately bugzilla doesn't have a great notion of "temporarily change the code in a given manner and track an experiment", but that's what we're doing here. 152395 19 eric 2009-10-05 11:38:34 -0700 Sounds good. I'm going to clear the r+ flags since our "to-be-committed" queue doesn't understand "temporary change in code to try an experiment" either. :( 152396 20 39586 eric 2009-10-05 11:39:08 -0700 Comment on attachment 39586 Slightly better preprocessor style this time Clearing darin's review on this obsolete patch. 152397 21 39681 eric 2009-10-05 11:39:25 -0700 Comment on attachment 39681 Don't fire unload events when leaving a page if the document has been cached. Clearing sam's review now that this patch has been landed. 158219 22 beidson 2009-10-27 09:43:11 -0700 We haven't seen a rash of bug reports in the WebKit project as we expected we might. But we do know: -Disabling this exemption cause a couple of benchmark performance regressions -Some important corporate intranet apps rely on the status-quo unload handler behavior, and this change breaks them. -Other browsers have seen substantial negative effects in analytics/marketshare/usage numbers when they fired unload handlers less often than the status quo. While the perf. regressions could be explored and fixed, then other two points are pretty damning. For now I think we're re-enabling this check. 158221 23 beidson 2009-10-27 09:45:18 -0700 Re-enabled in http://trac.webkit.org/projects/webkit/changeset/50153 I'll close this bug for now. If/when we come up with new ideas to improve this situation in the future, I'll file a new bug. 158226 24 jaffathecake 2009-10-27 10:00:55 -0700 What about my earlier comment about caching the state of the page before unload? If it's being ignored because the comment should have been made elsewhere, let me know. "We haven't seen a rash of bug reports in the WebKit project as we expected we might." I'm very surprised by this. The latest version of jQuery (1.3.2) detaches ALL event listeners on unload. This should have broken most interactivity on the page added via jQuery. I've just checked in the jQuery trunk, they've changed this code so it only runs in IE, however that isn't in an official jQuery release yet. "Other browsers have seen substantial negative effects in analytics/marketshare/usage numbers when they fired unload handlers less often than the status quo." Is this really a deal breaker? Shouldn't the priority be the user experience rather than usage stats? Could you work around this by sending an async request to the page after it's called back from the page cache? "Some important corporate intranet apps rely on the status-quo unload handler behavior, and this change breaks them." Do these also operate in standards mode? Either way, caching the state of the page before unload is called should solve this. Jake. 158240 25 beidson 2009-10-27 10:18:02 -0700 I thought I replied to your suggestion when you originally made it. Sometimes I fail at using bugzilla correctly ;) "You could save the page to the page cache, then fire the unload event and navigate away. If back is clicked, the page will be returned in the state it was BEFORE unload was fired." There's two major problems with this approach. 1 - It's technologically VERY very difficult. No, not impossible, and we're not afraid of doing difficult things, but this would require a major re-architecturing of both JavaScripteCore and WebCore to accomplish. A major undertaking that we'd need to conclusively know was worth it. 2 - Loads and unloads need to be balance. What you're suggesting opens the possibility that unload is fired twice (Or more!) for only a single corresponding load event. This is bad(tm) :) >Is this really a deal breaker? Shouldn't the priority be the user experience rather than usage stats? It's not only usage stats - it's ALL negative side effects from us not running code that programmers think we'll run. This whole experiment was about pushing the limits in balancing user experience and fulfilling an API contract with developers. We learned a lot. We raised awareness. This is good for users in the long run. >Could you work around this by sending an async request to the page after it's called back from the page cache? I encourage you to follow progress in https://bugs.webkit.org/show_bug.cgi?id=30457 and https://bugs.webkit.org/show_bug.cgi?id=30458. But without some heuristic manner of determining all of the side effects of running an unload handler (which we most certainly do not have) that still doesn't solve the page cache problem. 168248 26 klobag 2009-12-02 20:32:36 -0800 Brady, what happen if we call unload as before, which means reverting http://trac.webkit.org/changeset/48456, then we call m_document->dispatchWindowLoadEvent() in CachedFrameBase::restore(). This way, loadEvent and unloadEvent are balanced. Even unload may destroy some critical part, it is better than not cache at all. 168321 27 jaffathecake 2009-12-03 01:21:49 -0800 It'd be interesting to test, but I predict it would break pages. It's very rare for an unload handler to be the opposite of a load handler. Load tends to be used to modify elements on the page & start timers, unless people are reverting these things in unload, you'll start getting duplicates appearing on the page. 39585 2009-09-14 20:05:55 -0700 2009-09-14 20:14:18 -0700 Disable the unload handler check on Mac + Windows unloadHandler.txt text/plain 4376 beidson SW5kZXg6IFdlYkNvcmUvQ2hhbmdlTG9nCj09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT0KLS0tIFdlYkNvcmUvQ2hhbmdlTG9n CShyZXZpc2lvbiA0ODM4MykKKysrIFdlYkNvcmUvQ2hhbmdlTG9nCSh3b3JraW5nIGNvcHkpCkBA IC0xLDMgKzEsMTcgQEAKKzIwMDktMDktMTQgIEJyYWR5IEVpZHNvbiAgPGJlaWRzb25AYXBwbGUu Y29tPgorCisgICAgICAgIFJldmlld2VkIGJ5IE5PQk9EWSAoT09QUyEpLgorCisgICAgICAgIFBh Z2UgY2FjaGUgc2hvdWxkIHN1cHBvcnQgcGFnZXMgd2l0aCB1bmxvYWQgaGFuZGxlcnMgYXMgYW4g ZXhwZXJpbWVudC4KKyAgICAgICAgPHJkYXI6Ly9wcm9ibGVtLzcxOTY0ODU+IGFuZCBodHRwczov L2J1Z3Mud2Via2l0Lm9yZy9zaG93X2J1Zy5jZ2k/aWQ9MjkwMjEKKworICAgICAgICBGb3IgdGhl IGV4cGVyaW1lbnQsIG9ubHkgYWxsb3cgcGFnZXMgd2l0aCB1bmxvYWQgaGFuZGxlcnMgaW50byB0 aGUgUGFnZSBDYWNoZSBvbgorICAgICAgICBNYWMgYW5kIFdpbmRvd3MuCisKKyAgICAgICAgKiBs b2FkZXIvRnJhbWVMb2FkZXIuY3BwOgorICAgICAgICAoV2ViQ29yZTo6RnJhbWVMb2FkZXI6OmNh bkNhY2hlUGFnZUNvbnRhaW5pbmdUaGlzRnJhbWUpOgorICAgICAgICAoV2ViQ29yZTo6RnJhbWVM b2FkZXI6OmxvZ0NhbkNhY2hlRnJhbWVEZWNpc2lvbik6CisKIDIwMDktMDktMTQgIFNpbW9uIEZy YXNlciAgPHNpbW9uLmZyYXNlckBhcHBsZS5jb20+CiAKICAgICAgICAgRml4IHRoZSBjaGFuZ2Ug aW4gcjQ4Mzc2IHRvIGRvIGEgbnVsbC1jaGVjayB0byBmaXggYSBjcmFzaGluZyB0ZXN0LgpJbmRl eDogV2ViQ29yZS9sb2FkZXIvRnJhbWVMb2FkZXIuY3BwCj09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT0KLS0tIFdlYkNvcmUv bG9hZGVyL0ZyYW1lTG9hZGVyLmNwcAkocmV2aXNpb24gNDgzODApCisrKyBXZWJDb3JlL2xvYWRl ci9GcmFtZUxvYWRlci5jcHAJKHdvcmtpbmcgY29weSkKQEAgLTExMSw2ICsxMTEsMTAgQEAKICNp bmNsdWRlICJTVkdWaWV3U3BlYy5oIgogI2VuZGlmCiAKKyNpZiBQTEFURk9STShNQUMpIHx8IFBM QVRGT1JNKFdJTikKKyNkZWZpbmUgUEFHRV9DQUNIRV9BQ0NFUFRTX1VOTE9BRF9IQU5ETEVSUyAx CisjZW5kaWYKKwogbmFtZXNwYWNlIFdlYkNvcmUgewogCiAjaWYgRU5BQkxFKFNWRykKQEAgLTE4 NDgsNyArMTg1Miw5IEBAIGJvb2wgRnJhbWVMb2FkZXI6OmNhbkNhY2hlUGFnZUNvbnRhaW5pbmcK ICAgICAgICAgLy8gdGhlIHJpZ2h0IE5QT2JqZWN0cy4gU2VlIDxyZGFyOi8vcHJvYmxlbS81MTk3 MDQxPiBmb3IgbW9yZSBpbmZvcm1hdGlvbi4KICAgICAgICAgJiYgIW1fY29udGFpbnNQbHVnSW5z CiAgICAgICAgICYmICFtX1VSTC5wcm90b2NvbElzKCJodHRwcyIpCisjaWYgIWRlZmluZWQoUEFH RV9DQUNIRV9BQ0NFUFRTX1VOTE9BRF9IQU5ETEVSUykKICAgICAgICAgJiYgKCFtX2ZyYW1lLT5k b21XaW5kb3coKSB8fCAhbV9mcmFtZS0+ZG9tV2luZG93KCktPmhhc0V2ZW50TGlzdGVuZXIoZXZl bnROYW1lcygpLnVubG9hZEV2ZW50KSkKKyNlbmRpZgogI2lmIEVOQUJMRShEQVRBQkFTRSkKICAg ICAgICAgJiYgIW1fZnJhbWUtPmRvY3VtZW50KCktPmhhc09wZW5EYXRhYmFzZXMoKQogI2VuZGlm CkBAIC0xOTkzLDggKzE5OTksMTAgQEAgYm9vbCBGcmFtZUxvYWRlcjo6bG9nQ2FuQ2FjaGVGcmFt ZURlY2lzaQogICAgICAgICAgICAgeyBQQ0xPRygiICAgLUZyYW1lIGNvbnRhaW5zIHBsdWdpbnMi KTsgY2Fubm90Q2FjaGUgPSB0cnVlOyB9CiAgICAgICAgIGlmIChtX1VSTC5wcm90b2NvbElzKCJo dHRwcyIpKQogICAgICAgICAgICAgeyBQQ0xPRygiICAgLUZyYW1lIGlzIEhUVFBTIik7IGNhbm5v dENhY2hlID0gdHJ1ZTsgfQorI2lmICFkZWZpbmVkKFBBR0VfQ0FDSEVfQUNDRVBUU19VTkxPQURf SEFORExFUlMpCiAgICAgICAgIGlmIChtX2ZyYW1lLT5kb21XaW5kb3coKSAmJiBtX2ZyYW1lLT5k b21XaW5kb3coKS0+aGFzRXZlbnRMaXN0ZW5lcihldmVudE5hbWVzKCkudW5sb2FkRXZlbnQpKQog ICAgICAgICAgICAgeyBQQ0xPRygiICAgLUZyYW1lIGhhcyBhbiB1bmxvYWQgZXZlbnQgbGlzdGVu ZXIiKTsgY2Fubm90Q2FjaGUgPSB0cnVlOyB9CisjZW5kaWYKICNpZiBFTkFCTEUoREFUQUJBU0Up CiAgICAgICAgIGlmIChtX2ZyYW1lLT5kb2N1bWVudCgpLT5oYXNPcGVuRGF0YWJhc2VzKCkpCiAg ICAgICAgICAgICB7IFBDTE9HKCIgICAtRnJhbWUgaGFzIG9wZW4gZGF0YWJhc2UgaGFuZGxlcyIp OyBjYW5ub3RDYWNoZSA9IHRydWU7IH0KSW5kZXg6IExheW91dFRlc3RzL0NoYW5nZUxvZwo9PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09Ci0tLSBMYXlvdXRUZXN0cy9DaGFuZ2VMb2cJKHJldmlzaW9uIDQ4MzgzKQorKysgTGF5 b3V0VGVzdHMvQ2hhbmdlTG9nCSh3b3JraW5nIGNvcHkpCkBAIC0xLDMgKzEsMTYgQEAKKzIwMDkt MDktMTQgIEJyYWR5IEVpZHNvbiAgPGJlaWRzb25AYXBwbGUuY29tPgorCisgICAgICAgIFJldmll d2VkIGJ5IE5PQk9EWSAoT09QUyEpLgorCisgICAgICAgIFBhZ2UgY2FjaGUgc2hvdWxkIHN1cHBv cnQgcGFnZXMgd2l0aCB1bmxvYWQgaGFuZGxlcnMgYXMgYW4gZXhwZXJpbWVudC4KKyAgICAgICAg PHJkYXI6Ly9wcm9ibGVtLzcxOTY0ODU+IGFuZCBodHRwczovL2J1Z3Mud2Via2l0Lm9yZy9zaG93 X2J1Zy5jZ2k/aWQ9MjkwMjEKKworICAgICAgICBEaXNhYmxlIGZhc3QvbG9hZGVyL2ZyYW1lcy13 aXRoLXVubG9hZC1oYW5kbGVycy1pbi1wYWdlLWNhY2hlLmh0bWwgYXMgaXQgZG9lc24ndAorICAg ICAgICBtYWtlIHNlbnNlIGluIHRoZSBjb250ZXh0IG9mIHRoaXMgZXhwZXJpbWVudC4KKworICAg ICAgICAqIHBsYXRmb3JtL21hYy9Ta2lwcGVkOiAKKyAgICAgICAgKiBwbGF0Zm9ybS93aW4vU2tp cHBlZDoKKwogMjAwOS0wOS0xNCAgQ2hyaXMgTWFycmluICA8Y21hcnJpbkBhcHBsZS5jb20+CiAK ICAgICAgICAgUmV2aWV3ZWQgYnkgT2xpdmVyIEh1bnQuCkluZGV4OiBMYXlvdXRUZXN0cy9wbGF0 Zm9ybS9tYWMvU2tpcHBlZAo9PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09Ci0tLSBMYXlvdXRUZXN0cy9wbGF0Zm9ybS9tYWMv U2tpcHBlZAkocmV2aXNpb24gNDgzODApCisrKyBMYXlvdXRUZXN0cy9wbGF0Zm9ybS9tYWMvU2tp cHBlZAkod29ya2luZyBjb3B5KQpAQCAtMTAxLDMgKzEwMSw4IEBAIGZhc3QvbGF5ZXJzL3ZpZGVv LWxheWVyLmh0bWwKIG1lZGlhL2F1ZGlvLWRlbGV0ZS13aGlsZS1zdGVwLWJ1dHRvbi1jbGlja2Vk Lmh0bWwKIG1lZGlhL3ZpZGVvLWNvbnRyb2xzLXRyYW5zZm9ybWVkLmh0bWwKIG1lZGlhL3ZpZGVv LWNvbnRyb2xzLXpvb21lZC5odG1sCisKKyMgQXMgbG9uZyBhcyBwYWdlcyB3aXRoIHVubG9hZCBo YW5kbGVycyBhcmUgcHV0IGluIHRoZSBQYWdlIENhY2hlLCB0aGlzIHRlc3QgZG9lc24ndCBtYWtl IGFueSBzZW5zZSBmb3IgTWFjLgorIyA8cmRhcjovL3Byb2JsZW0vNzE5NjQ4NT4gYW5kIGh0dHBz Oi8vYnVncy53ZWJraXQub3JnL3Nob3dfYnVnLmNnaT9pZD0yOTAyMQorZmFzdC9sb2FkZXIvZnJh bWVzLXdpdGgtdW5sb2FkLWhhbmRsZXJzLWluLXBhZ2UtY2FjaGUuaHRtbAorCkluZGV4OiBMYXlv dXRUZXN0cy9wbGF0Zm9ybS93aW4vU2tpcHBlZAo9PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09Ci0tLSBMYXlvdXRUZXN0cy9w bGF0Zm9ybS93aW4vU2tpcHBlZAkocmV2aXNpb24gNDgzODApCisrKyBMYXlvdXRUZXN0cy9wbGF0 Zm9ybS93aW4vU2tpcHBlZAkod29ya2luZyBjb3B5KQpAQCAtNjYyLDMgKzY2Miw2IEBAIG1lZGlh L3ZpZGVvLWNvbnRyb2xzLXpvb21lZC5odG1sCiAjIE1peGVkIGNvbnRlbnQgdGVzdHMgZmFpbGlu ZyBvbiBXaW5kb3dzIDxodHRwOi8vd2Via2l0Lm9yZy9iLzI5MjEzPgogaHR0cC90ZXN0cy9zZWN1 cml0eS9taXhlZENvbnRlbnQKIAorIyBBcyBsb25nIGFzIHBhZ2VzIHdpdGggdW5sb2FkIGhhbmRs ZXJzIGFyZSBwdXQgaW4gdGhlIFBhZ2UgQ2FjaGUsIHRoaXMgdGVzdCBkb2Vzbid0IG1ha2UgYW55 IHNlbnNlIGZvciBXaW5kb3dzLgorIyA8cmRhcjovL3Byb2JsZW0vNzE5NjQ4NT4gYW5kIGh0dHBz Oi8vYnVncy53ZWJraXQub3JnL3Nob3dfYnVnLmNnaT9pZD0yOTAyMQorZmFzdC9sb2FkZXIvZnJh bWVzLXdpdGgtdW5sb2FkLWhhbmRsZXJzLWluLXBhZ2UtY2FjaGUuaHRtbAo= 39586 2009-09-14 20:14:18 -0700 2009-10-05 11:39:08 -0700 Slightly better preprocessor style this time unloadHandler.txt text/plain 4362 beidson SW5kZXg6IFdlYkNvcmUvQ2hhbmdlTG9nCj09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT0KLS0tIFdlYkNvcmUvQ2hhbmdlTG9n CShyZXZpc2lvbiA0ODM4MykKKysrIFdlYkNvcmUvQ2hhbmdlTG9nCSh3b3JraW5nIGNvcHkpCkBA IC0xLDMgKzEsMTcgQEAKKzIwMDktMDktMTQgIEJyYWR5IEVpZHNvbiAgPGJlaWRzb25AYXBwbGUu Y29tPgorCisgICAgICAgIFJldmlld2VkIGJ5IE5PQk9EWSAoT09QUyEpLgorCisgICAgICAgIFBh Z2UgY2FjaGUgc2hvdWxkIHN1cHBvcnQgcGFnZXMgd2l0aCB1bmxvYWQgaGFuZGxlcnMgYXMgYW4g ZXhwZXJpbWVudC4KKyAgICAgICAgPHJkYXI6Ly9wcm9ibGVtLzcxOTY0ODU+IGFuZCBodHRwczov L2J1Z3Mud2Via2l0Lm9yZy9zaG93X2J1Zy5jZ2k/aWQ9MjkwMjEKKworICAgICAgICBGb3IgdGhl IGV4cGVyaW1lbnQsIG9ubHkgYWxsb3cgcGFnZXMgd2l0aCB1bmxvYWQgaGFuZGxlcnMgaW50byB0 aGUgUGFnZSBDYWNoZSBvbgorICAgICAgICBNYWMgYW5kIFdpbmRvd3MuCisKKyAgICAgICAgKiBs b2FkZXIvRnJhbWVMb2FkZXIuY3BwOgorICAgICAgICAoV2ViQ29yZTo6RnJhbWVMb2FkZXI6OmNh bkNhY2hlUGFnZUNvbnRhaW5pbmdUaGlzRnJhbWUpOgorICAgICAgICAoV2ViQ29yZTo6RnJhbWVM b2FkZXI6OmxvZ0NhbkNhY2hlRnJhbWVEZWNpc2lvbik6CisKIDIwMDktMDktMTQgIFNpbW9uIEZy YXNlciAgPHNpbW9uLmZyYXNlckBhcHBsZS5jb20+CiAKICAgICAgICAgRml4IHRoZSBjaGFuZ2Ug aW4gcjQ4Mzc2IHRvIGRvIGEgbnVsbC1jaGVjayB0byBmaXggYSBjcmFzaGluZyB0ZXN0LgpJbmRl eDogV2ViQ29yZS9sb2FkZXIvRnJhbWVMb2FkZXIuY3BwCj09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT0KLS0tIFdlYkNvcmUv bG9hZGVyL0ZyYW1lTG9hZGVyLmNwcAkocmV2aXNpb24gNDgzODApCisrKyBXZWJDb3JlL2xvYWRl ci9GcmFtZUxvYWRlci5jcHAJKHdvcmtpbmcgY29weSkKQEAgLTExMSw2ICsxMTEsMTAgQEAKICNp bmNsdWRlICJTVkdWaWV3U3BlYy5oIgogI2VuZGlmCiAKKyNpZiBQTEFURk9STShNQUMpIHx8IFBM QVRGT1JNKFdJTikKKyNkZWZpbmUgUEFHRV9DQUNIRV9BQ0NFUFRTX1VOTE9BRF9IQU5ETEVSUwor I2VuZGlmCisKIG5hbWVzcGFjZSBXZWJDb3JlIHsKIAogI2lmIEVOQUJMRShTVkcpCkBAIC0xODQ4 LDcgKzE4NTIsOSBAQCBib29sIEZyYW1lTG9hZGVyOjpjYW5DYWNoZVBhZ2VDb250YWluaW5nCiAg ICAgICAgIC8vIHRoZSByaWdodCBOUE9iamVjdHMuIFNlZSA8cmRhcjovL3Byb2JsZW0vNTE5NzA0 MT4gZm9yIG1vcmUgaW5mb3JtYXRpb24uCiAgICAgICAgICYmICFtX2NvbnRhaW5zUGx1Z0lucwog ICAgICAgICAmJiAhbV9VUkwucHJvdG9jb2xJcygiaHR0cHMiKQorI2lmbmRlZiBQQUdFX0NBQ0hF X0FDQ0VQVFNfVU5MT0FEX0hBTkRMRVJTCiAgICAgICAgICYmICghbV9mcmFtZS0+ZG9tV2luZG93 KCkgfHwgIW1fZnJhbWUtPmRvbVdpbmRvdygpLT5oYXNFdmVudExpc3RlbmVyKGV2ZW50TmFtZXMo KS51bmxvYWRFdmVudCkpCisjZW5kaWYKICNpZiBFTkFCTEUoREFUQUJBU0UpCiAgICAgICAgICYm ICFtX2ZyYW1lLT5kb2N1bWVudCgpLT5oYXNPcGVuRGF0YWJhc2VzKCkKICNlbmRpZgpAQCAtMTk5 Myw4ICsxOTk5LDEwIEBAIGJvb2wgRnJhbWVMb2FkZXI6OmxvZ0NhbkNhY2hlRnJhbWVEZWNpc2kK ICAgICAgICAgICAgIHsgUENMT0coIiAgIC1GcmFtZSBjb250YWlucyBwbHVnaW5zIik7IGNhbm5v dENhY2hlID0gdHJ1ZTsgfQogICAgICAgICBpZiAobV9VUkwucHJvdG9jb2xJcygiaHR0cHMiKSkK ICAgICAgICAgICAgIHsgUENMT0coIiAgIC1GcmFtZSBpcyBIVFRQUyIpOyBjYW5ub3RDYWNoZSA9 IHRydWU7IH0KKyNpZm5kZWYgUEFHRV9DQUNIRV9BQ0NFUFRTX1VOTE9BRF9IQU5ETEVSUwogICAg ICAgICBpZiAobV9mcmFtZS0+ZG9tV2luZG93KCkgJiYgbV9mcmFtZS0+ZG9tV2luZG93KCktPmhh c0V2ZW50TGlzdGVuZXIoZXZlbnROYW1lcygpLnVubG9hZEV2ZW50KSkKICAgICAgICAgICAgIHsg UENMT0coIiAgIC1GcmFtZSBoYXMgYW4gdW5sb2FkIGV2ZW50IGxpc3RlbmVyIik7IGNhbm5vdENh Y2hlID0gdHJ1ZTsgfQorI2VuZGlmCiAjaWYgRU5BQkxFKERBVEFCQVNFKQogICAgICAgICBpZiAo bV9mcmFtZS0+ZG9jdW1lbnQoKS0+aGFzT3BlbkRhdGFiYXNlcygpKQogICAgICAgICAgICAgeyBQ Q0xPRygiICAgLUZyYW1lIGhhcyBvcGVuIGRhdGFiYXNlIGhhbmRsZXMiKTsgY2Fubm90Q2FjaGUg PSB0cnVlOyB9CkluZGV4OiBMYXlvdXRUZXN0cy9DaGFuZ2VMb2cKPT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PQotLS0gTGF5 b3V0VGVzdHMvQ2hhbmdlTG9nCShyZXZpc2lvbiA0ODM4MykKKysrIExheW91dFRlc3RzL0NoYW5n ZUxvZwkod29ya2luZyBjb3B5KQpAQCAtMSwzICsxLDE2IEBACisyMDA5LTA5LTE0ICBCcmFkeSBF aWRzb24gIDxiZWlkc29uQGFwcGxlLmNvbT4KKworICAgICAgICBSZXZpZXdlZCBieSBOT0JPRFkg KE9PUFMhKS4KKworICAgICAgICBQYWdlIGNhY2hlIHNob3VsZCBzdXBwb3J0IHBhZ2VzIHdpdGgg dW5sb2FkIGhhbmRsZXJzIGFzIGFuIGV4cGVyaW1lbnQuCisgICAgICAgIDxyZGFyOi8vcHJvYmxl bS83MTk2NDg1PiBhbmQgaHR0cHM6Ly9idWdzLndlYmtpdC5vcmcvc2hvd19idWcuY2dpP2lkPTI5 MDIxCisKKyAgICAgICAgRGlzYWJsZSBmYXN0L2xvYWRlci9mcmFtZXMtd2l0aC11bmxvYWQtaGFu ZGxlcnMtaW4tcGFnZS1jYWNoZS5odG1sIGFzIGl0IGRvZXNuJ3QKKyAgICAgICAgbWFrZSBzZW5z ZSBpbiB0aGUgY29udGV4dCBvZiB0aGlzIGV4cGVyaW1lbnQuCisKKyAgICAgICAgKiBwbGF0Zm9y bS9tYWMvU2tpcHBlZDogCisgICAgICAgICogcGxhdGZvcm0vd2luL1NraXBwZWQ6CisKIDIwMDkt MDktMTQgIENocmlzIE1hcnJpbiAgPGNtYXJyaW5AYXBwbGUuY29tPgogCiAgICAgICAgIFJldmll d2VkIGJ5IE9saXZlciBIdW50LgpJbmRleDogTGF5b3V0VGVzdHMvcGxhdGZvcm0vbWFjL1NraXBw ZWQKPT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PQotLS0gTGF5b3V0VGVzdHMvcGxhdGZvcm0vbWFjL1NraXBwZWQJKHJldmlz aW9uIDQ4MzgwKQorKysgTGF5b3V0VGVzdHMvcGxhdGZvcm0vbWFjL1NraXBwZWQJKHdvcmtpbmcg Y29weSkKQEAgLTEwMSwzICsxMDEsOCBAQCBmYXN0L2xheWVycy92aWRlby1sYXllci5odG1sCiBt ZWRpYS9hdWRpby1kZWxldGUtd2hpbGUtc3RlcC1idXR0b24tY2xpY2tlZC5odG1sCiBtZWRpYS92 aWRlby1jb250cm9scy10cmFuc2Zvcm1lZC5odG1sCiBtZWRpYS92aWRlby1jb250cm9scy16b29t ZWQuaHRtbAorCisjIEFzIGxvbmcgYXMgcGFnZXMgd2l0aCB1bmxvYWQgaGFuZGxlcnMgYXJlIHB1 dCBpbiB0aGUgUGFnZSBDYWNoZSwgdGhpcyB0ZXN0IGRvZXNuJ3QgbWFrZSBhbnkgc2Vuc2UgZm9y IE1hYy4KKyMgPHJkYXI6Ly9wcm9ibGVtLzcxOTY0ODU+IGFuZCBodHRwczovL2J1Z3Mud2Via2l0 Lm9yZy9zaG93X2J1Zy5jZ2k/aWQ9MjkwMjEKK2Zhc3QvbG9hZGVyL2ZyYW1lcy13aXRoLXVubG9h ZC1oYW5kbGVycy1pbi1wYWdlLWNhY2hlLmh0bWwKKwpJbmRleDogTGF5b3V0VGVzdHMvcGxhdGZv cm0vd2luL1NraXBwZWQKPT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PQotLS0gTGF5b3V0VGVzdHMvcGxhdGZvcm0vd2luL1Nr aXBwZWQJKHJldmlzaW9uIDQ4MzgwKQorKysgTGF5b3V0VGVzdHMvcGxhdGZvcm0vd2luL1NraXBw ZWQJKHdvcmtpbmcgY29weSkKQEAgLTY2MiwzICs2NjIsNiBAQCBtZWRpYS92aWRlby1jb250cm9s cy16b29tZWQuaHRtbAogIyBNaXhlZCBjb250ZW50IHRlc3RzIGZhaWxpbmcgb24gV2luZG93cyA8 aHR0cDovL3dlYmtpdC5vcmcvYi8yOTIxMz4KIGh0dHAvdGVzdHMvc2VjdXJpdHkvbWl4ZWRDb250 ZW50CiAKKyMgQXMgbG9uZyBhcyBwYWdlcyB3aXRoIHVubG9hZCBoYW5kbGVycyBhcmUgcHV0IGlu IHRoZSBQYWdlIENhY2hlLCB0aGlzIHRlc3QgZG9lc24ndCBtYWtlIGFueSBzZW5zZSBmb3IgV2lu ZG93cy4KKyMgPHJkYXI6Ly9wcm9ibGVtLzcxOTY0ODU+IGFuZCBodHRwczovL2J1Z3Mud2Via2l0 Lm9yZy9zaG93X2J1Zy5jZ2k/aWQ9MjkwMjEKK2Zhc3QvbG9hZGVyL2ZyYW1lcy13aXRoLXVubG9h ZC1oYW5kbGVycy1pbi1wYWdlLWNhY2hlLmh0bWwK 39681 2009-09-16 23:17:15 -0700 2009-10-05 11:39:25 -0700 Don't fire unload events when leaving a page if the document has been cached. patch.txt text/plain 1549 beidson SW5kZXg6IFdlYkNvcmUvQ2hhbmdlTG9nCj09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT0KLS0tIFdlYkNvcmUvQ2hhbmdlTG9n CShyZXZpc2lvbiA0ODQ1NCkKKysrIFdlYkNvcmUvQ2hhbmdlTG9nCSh3b3JraW5nIGNvcHkpCkBA IC0xLDMgKzEsMTUgQEAKKzIwMDktMDktMTYgIEJyYWR5IEVpZHNvbiAgPGJlaWRzb25AYXBwbGUu Y29tPgorCisgICAgICAgIFJldmlld2VkIGJ5IE5PQk9EWSAoT09QUyEpLgorCisgICAgICAgIEV4 cGxvcmUgYWxsb3dpbmcgcGFnZXMgd2l0aCB1bmxvYWQgaGFuZGxlcnMgaW50byB0aGUgUGFnZSBD YWNoZQorICAgICAgICBodHRwczovL2J1Z3Mud2Via2l0Lm9yZy9zaG93X2J1Zy5jZ2k/aWQ9Mjkw MjEKKworICAgICAgICBObyBuZXcgdGVzdHMuIChBbGwgcHJldmlvdXMgdGVzdHMgY29udGludWUg dG8gcGFzcykKKworICAgICAgICAqIGxvYWRlci9GcmFtZUxvYWRlci5jcHA6CisgICAgICAgIChX ZWJDb3JlOjpGcmFtZUxvYWRlcjo6c3RvcExvYWRpbmcpOiBJZiB0aGUgZG9jdW1lbnQgaXMgaW4g dGhlIHBhZ2UgY2FjaGUsIGRvbid0IGZpcmUgdGhlIHVubG9hZCBldmVudC4KKwogMjAwOS0wOS0x NiAgU2FtIFdlaW5pZyAgPHNhbUB3ZWJraXQub3JnPgogCiAgICAgICAgIFJldmlld2VkIGJ5IEFu ZGVycyBDYXJsc3Nvbi4KSW5kZXg6IFdlYkNvcmUvbG9hZGVyL0ZyYW1lTG9hZGVyLmNwcAo9PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09Ci0tLSBXZWJDb3JlL2xvYWRlci9GcmFtZUxvYWRlci5jcHAJKHJldmlzaW9uIDQ4NDUz KQorKysgV2ViQ29yZS9sb2FkZXIvRnJhbWVMb2FkZXIuY3BwCSh3b3JraW5nIGNvcHkpCkBAIC01 ODIsNyArNTgyLDggQEAgdm9pZCBGcmFtZUxvYWRlcjo6c3RvcExvYWRpbmcoVW5sb2FkRXZlbgog ICAgICAgICAgICAgICAgIGlmIChtX2ZyYW1lLT5kb21XaW5kb3coKSkgewogICAgICAgICAgICAg ICAgICAgICBpZiAodW5sb2FkRXZlbnRQb2xpY3kgPT0gVW5sb2FkRXZlbnRQb2xpY3lVbmxvYWRB bmRQYWdlSGlkZSkKICAgICAgICAgICAgICAgICAgICAgICAgIG1fZnJhbWUtPmRvbVdpbmRvdygp LT5kaXNwYXRjaFBhZ2VUcmFuc2l0aW9uRXZlbnQoRXZlbnROYW1lcygpLnBhZ2VoaWRlRXZlbnQs IG1fZnJhbWUtPmRvY3VtZW50KCktPmluUGFnZUNhY2hlKCkpOwotICAgICAgICAgICAgICAgICAg ICBtX2ZyYW1lLT5kb21XaW5kb3coKS0+ZGlzcGF0Y2hVbmxvYWRFdmVudCgpOworICAgICAgICAg ICAgICAgICAgICBpZiAoIW1fZnJhbWUtPmRvY3VtZW50KCktPmluUGFnZUNhY2hlKCkpCisgICAg ICAgICAgICAgICAgICAgICAgICBtX2ZyYW1lLT5kb21XaW5kb3coKS0+ZGlzcGF0Y2hVbmxvYWRF dmVudCgpOwogICAgICAgICAgICAgICAgIH0KICAgICAgICAgICAgICAgICBtX3VubG9hZEV2ZW50 QmVpbmdEaXNwYXRjaGVkID0gZmFsc2U7CiAgICAgICAgICAgICAgICAgaWYgKG1fZnJhbWUtPmRv Y3VtZW50KCkpCg==