33269 2010-01-06 13:53:35 -0800 Improve HTMLElement::tagPriority() 2010-01-07 20:24:50 -0800 1 1 1 Unclassified WebKit DOM 528+ (Nightly build) All All RESOLVED FIXED P3 Minor --- 1 tkent webkit-unassigned darin mjs webkit.review.bot oldest_to_newest 177744 0 tkent 2010-01-06 13:53:35 -0800 HTMLElement::tagPriority() compares AtomicStringImpl pointers 18 times for a generic element such as <span>. 177749 1 45992 tkent 2010-01-06 13:58:15 -0800 Created attachment 45992 Proposed patch 177752 2 webkit.review.bot 2010-01-06 13:58:55 -0800 style-queue ran check-webkit-style on attachment 45992 without any errors. 177758 3 45992 darin 2010-01-06 14:08:12 -0800 Comment on attachment 45992 Proposed patch > +struct Empty1IntHashTraits : HashTraits<int> { This seems a slightly heavyweight wait of making this use a map. I wonder if there's some other clever approach that avoids custom hash traits. One thing we could do is this: 1) Increase all tag priorities by 1 in all the code. 2) Use find and return a tag priority of 2 if we don't find anything, instead of using get. I guess your way is probably better. > + static const bool emptyValueIsZero = false; > + static const bool needsDestruction = false; > + static int emptyValue() { return 1; } > + static void constructDeletedValue(int& slot) { slot = -1; } > + static bool isDeletedValue(int value) { return value == -1; } There's no need for you to override needsDestruction, constructDeletedValue, or isDeletedValue. You can just use the inherited versions of those from the base class. > + DEFINE_STATIC_LOCAL(TagPriorityMap, tagPriorityMap, ()); > + if (tagPriorityMap.isEmpty()) { > + tagPriorityMap.add(wbrTag.localName().impl(), 0); > + > + tagPriorityMap.add(addressTag.localName().impl(), 3); > + tagPriorityMap.add(ddTag.localName().impl(), 3); > + tagPriorityMap.add(dtTag.localName().impl(), 3); > + tagPriorityMap.add(noscriptTag.localName().impl(), 3); > + tagPriorityMap.add(rpTag.localName().impl(), 3); > + tagPriorityMap.add(rtTag.localName().impl(), 3); > + > + // 5 is same as <div>'s priority. > + tagPriorityMap.add(articleTag.localName().impl(), 5); > + tagPriorityMap.add(asideTag.localName().impl(), 5); > + tagPriorityMap.add(centerTag.localName().impl(), 5); > + tagPriorityMap.add(footerTag.localName().impl(), 5); > + tagPriorityMap.add(headerTag.localName().impl(), 5); > + tagPriorityMap.add(nobrTag.localName().impl(), 5); > + tagPriorityMap.add(rubyTag.localName().impl(), 5); > + tagPriorityMap.add(navTag.localName().impl(), 5); > + tagPriorityMap.add(sectionTag.localName().impl(), 5); > + > + tagPriorityMap.add(noembedTag.localName().impl(), 10); > + tagPriorityMap.add(noframesTag.localName().impl(), 10); > + > + // Return 1 for other tags. It's same as <span>. > + // This way custom tag name elements will behave like inline spans. > + } > > - // Same values as <span>. This way custom tag name elements will behave like inline spans. > - return 1; > + return tagPriorityMap.get(localName().impl()); > } I think it would be best to put the code to initialize the map into a separate function, initializeTagPriorityMap, that takes the map as an argument. That will make the tagPriorityMap function easier to read and even ever so slightly more efficient since it will be smaller with better code locality. Is there a guarantee that localName().impl() is never 0? review- so you can at least take out the unneeded overrides 177801 4 45999 tkent 2010-01-06 15:02:36 -0800 Created attachment 45999 Proposed patch (rev.2) (In reply to comment #3) > (From update of attachment 45992 [details]) > > + static const bool emptyValueIsZero = false; > > + static const bool needsDestruction = false; > > + static int emptyValue() { return 1; } > > + static void constructDeletedValue(int& slot) { slot = -1; } > > + static bool isDeletedValue(int value) { return value == -1; } > > There's no need for you to override needsDestruction, constructDeletedValue, or > isDeletedValue. You can just use the inherited versions of those from the base > class. ok, I removed them. > I think it would be best to put the code to initialize the map into a separate > function, initializeTagPriorityMap, that takes the map as an argument. That > will make the tagPriorityMap function easier to read and even ever so slightly > more efficient since it will be smaller with better code locality. Done. > Is there a guarantee that localName().impl() is never 0? I think so for the current code though there is no ASSERTION for it. However, 0 makes no problems because the HashMap key is AtomicStringImpl*. The generic PtrHash is used and tagPriorityMap.get(0) simply returns emptyValue(). 177803 5 webkit.review.bot 2010-01-06 15:03:41 -0800 style-queue ran check-webkit-style on attachment 45999 without any errors. 177844 6 darin 2010-01-06 16:15:28 -0800 (In reply to comment #4) > (In reply to comment #3) > > Is there a guarantee that localName().impl() is never 0? > > I think so for the current code though there is no ASSERTION for it. > However, 0 makes no problems because the HashMap key is AtomicStringImpl*. The > generic PtrHash is used and tagPriorityMap.get(0) simply returns emptyValue(). That's not true. If you call tagPriorityMap.get(0) there’s a chance that you could get an uninitialized value from an empty hash table entry. There is no special check for the empty value and if the value you pass hashes to a slot that has a 0 in it, then it will return the value in that slot. It is *not* safe to call get(0). 177850 7 45999 darin 2010-01-06 16:26:36 -0800 Comment on attachment 45999 Proposed patch (rev.2) > +// TagPriorityMap returns 1 for unregistered tags. It's same as <span>. > +// This way custom tag name elements will behave like inline spans. I think this comment should be at the bottom of initializeTagPriorityMap. > +static const TagPriorityMap* initializeTagPriorityMap() Given that this actually creates the map too, I suggest naming it createTagPriorityMap. r=me assuming that localName().impl() is *not* ever 0 If it can be zero, then this patch is not OK. 177911 8 tkent 2010-01-06 18:30:42 -0800 I have confirmed that impl() never be null, and this is off-topic. But I'd like to understand this. (In reply to comment #6) > > However, 0 makes no problems because the HashMap key is AtomicStringImpl*. The > > generic PtrHash is used and tagPriorityMap.get(0) simply returns emptyValue(). > > That's not true. > > If you call tagPriorityMap.get(0) there’s a chance that you could get an > uninitialized value from an empty hash table entry. There is no special check > for the empty value and if the value you pass hashes to a slot that has a 0 in > it, then it will return the value in that slot. > > It is *not* safe to call get(0). I read the code of HashTable.h and HashMap.h, and still wonder why it is not safe. You are talking about a case of empty HashTable, and it's safe for non-empty HashTable, right? In the empty case, HashTable::m_table is 0 and HashTable::lookup() seems to return (Value*)0 correctly for any keys. HashMap::get() returns emptyValue() for (Value*)0. I couldn't find a case that HashMap::get() returns uninitialized value. 177913 9 tkent 2010-01-06 18:35:14 -0800 (In reply to comment #7) > (From update of attachment 45999 [details]) > > +// TagPriorityMap returns 1 for unregistered tags. It's same as <span>. > > +// This way custom tag name elements will behave like inline spans. > > I think this comment should be at the bottom of initializeTagPriorityMap. Moved it. > > > +static const TagPriorityMap* initializeTagPriorityMap() > > Given that this actually creates the map too, I suggest naming it > createTagPriorityMap. Renamed it. > r=me assuming that localName().impl() is *not* ever 0 > > If it can be zero, then this patch is not OK. I read the HTMLParser, HTMLTokenizer, HTMLDocument, Document classes carefully and confirmed localName() never be null. I also confirmed all LayoutTests passed with ASSERT(localName().impl()). I'll add this assertion and will commit it. 177956 10 tkent 2010-01-06 20:47:49 -0800 Landed as r52899 <http://trac.webkit.org/changeset/52899> 178230 11 darin 2010-01-07 09:39:10 -0800 (In reply to comment #8) > I read the code of HashTable.h and HashMap.h, and still wonder why it is not > safe. > You are talking about a case of empty HashTable, and it's safe for non-empty > HashTable, right? I am talking about a non-empty map. HashTable::lookup will hash the empty key and get a hash value of 0, then the loop will look at the 0'th bucket in the hash table and if it is an empty bucket, will return a pointer to it. I believe the empty bucket is guaranteed to have a key of zero, but could easily have a value that's non-zero. That's how I understand the code. There may be something wrong with my analysis. In general, I have long believed that both the empty and deleted values are illegal as arguments to the HashTable::lookup function. I believe we need assertions to catch cases like that to make correct programming with this class template easier. 178500 12 tkent 2010-01-07 20:24:50 -0800 (In reply to comment #11) I catch you. HashTraits<P*> has emptyValueisZero=true, and HashTable::checkKey() called by HashTable::lookup() has ASSERT(!HashTranslator::equal(KeyTraits::emptyValue(), key)); It's enough for the reason of the wrong HashMap::get(0). Thank you! 45992 2010-01-06 13:58:15 -0800 2010-01-06 15:02:36 -0800 Proposed patch 0001-element-hash-1.patch text/plain 4241 tkent RnJvbSA0YjI4ZWFlNjk0ZTEwNWU1MDYyYmZkYzcxMTY0ZTI2OGUxNGJmM2VhIE1vbiBTZXAgMTcg MDA6MDA6MDAgMjAwMQpGcm9tOiBLZW50IFRhbXVyYSA8dGtlbnRAY2hyb21pdW0ub3JnPgpEYXRl OiBUaHUsIDcgSmFuIDIwMTAgMDY6NDk6NTYgKzA5MDAKU3ViamVjdDogW1BBVENIXSBlbGVtZW50 LWhhc2gtMQoKLS0tCiBXZWJDb3JlL0NoYW5nZUxvZyAgICAgICAgICAgIHwgICAxNiArKysrKysr KysrKysrKwogV2ViQ29yZS9odG1sL0hUTUxFbGVtZW50LmNwcCB8ICAgNDggKysrKysrKysrKysr KysrKysrKysrKysrKysrKysrKysrLS0tLS0tLS0KIDIgZmlsZXMgY2hhbmdlZCwgNTQgaW5zZXJ0 aW9ucygrKSwgMTAgZGVsZXRpb25zKC0pCgpkaWZmIC0tZ2l0IGEvV2ViQ29yZS9DaGFuZ2VMb2cg Yi9XZWJDb3JlL0NoYW5nZUxvZwppbmRleCAyNTViZDU4Li44MWY5MGJkIDEwMDY0NAotLS0gYS9X ZWJDb3JlL0NoYW5nZUxvZworKysgYi9XZWJDb3JlL0NoYW5nZUxvZwpAQCAtMSwzICsxLDE5IEBA CisyMDEwLTAxLTA2ICBLZW50IFRhbXVyYSAgPHRrZW50QGNocm9taXVtLm9yZz4KKworICAgICAg ICBSZXZpZXdlZCBieSBOT0JPRFkgKE9PUFMhKS4KKworICAgICAgICBVc2UgYSBzdGF0aWMgSGFz aE1hcCBmb3IgSFRNTEVsZW1lbnQ6OnRhZ1ByaW9yaXR5KCkuCisgICAgICAgIGh0dHBzOi8vYnVn cy53ZWJraXQub3JnL3Nob3dfYnVnLmNnaT9pZD0zMzI2OQorCisgICAgICAgIFRoZSBwcmlvciBj b2RlIGNvbXBhcmVzIEF0b21pY1N0cmluZ0ltcGwgcG9pbnRlcnMgMTggdGltZXMgYXQKKyAgICAg ICAgd29yc3QuIFRoaXMgY2hhbmdlIGF2b2lkcyBpdC4KKworICAgICAgICBObyBuZXcgdGVzdHMg YmVjYXVzZSB0aGlzIGlzIGp1c3QgYSByZWZhY3RvcmluZy4KKworICAgICAgICAqIGh0bWwvSFRN TEVsZW1lbnQuY3BwOgorICAgICAgICAoV2ViQ29yZTo6RW1wdHkxSW50SGFzaFRyYWl0cyk6IEEg SGFzaFRyYWl0cyB0byByZXR1cm4gMSBhcyB0aGUgZW1wdHkgdmFsdWUuCisgICAgICAgIChXZWJD b3JlOjpIVE1MRWxlbWVudDo6dGFnUHJpb3JpdHkpOiBVc2UgYSBzdGF0aWMgSGFzaE1hcC4KKwog MjAxMC0wMS0wNiAgU2ltb24gSGF1c21hbm4gIDxzaW1vbi5oYXVzbWFubkBub2tpYS5jb20+CiAK ICAgICAgICAgVW5yZXZpZXdlZCB0cml2aWFsIFF0IGJ1aWxkIGZpeC4KZGlmZiAtLWdpdCBhL1dl YkNvcmUvaHRtbC9IVE1MRWxlbWVudC5jcHAgYi9XZWJDb3JlL2h0bWwvSFRNTEVsZW1lbnQuY3Bw CmluZGV4IGZmYTBlZWEuLjE1MzczMDMgMTAwNjQ0Ci0tLSBhL1dlYkNvcmUvaHRtbC9IVE1MRWxl bWVudC5jcHAKKysrIGIvV2ViQ29yZS9odG1sL0hUTUxFbGVtZW50LmNwcApAQCAtODIsMTkgKzgy LDQ3IEBAIEhUTUxUYWdTdGF0dXMgSFRNTEVsZW1lbnQ6OmVuZFRhZ1JlcXVpcmVtZW50KCkgY29u c3QKICAgICByZXR1cm4gVGFnU3RhdHVzUmVxdWlyZWQ7CiB9CiAKK3N0cnVjdCBFbXB0eTFJbnRI YXNoVHJhaXRzIDogSGFzaFRyYWl0czxpbnQ+IHsKKyAgICBzdGF0aWMgY29uc3QgYm9vbCBlbXB0 eVZhbHVlSXNaZXJvID0gZmFsc2U7CisgICAgc3RhdGljIGNvbnN0IGJvb2wgbmVlZHNEZXN0cnVj dGlvbiA9IGZhbHNlOworICAgIHN0YXRpYyBpbnQgZW1wdHlWYWx1ZSgpIHsgcmV0dXJuIDE7IH0K KyAgICBzdGF0aWMgdm9pZCBjb25zdHJ1Y3REZWxldGVkVmFsdWUoaW50JiBzbG90KSB7IHNsb3Qg PSAtMTsgfQorICAgIHN0YXRpYyBib29sIGlzRGVsZXRlZFZhbHVlKGludCB2YWx1ZSkgeyByZXR1 cm4gdmFsdWUgPT0gLTE7IH0KK307Cit0eXBlZGVmIEhhc2hNYXA8QXRvbWljU3RyaW5nSW1wbCos IGludCwgRGVmYXVsdEhhc2g8QXRvbWljU3RyaW5nSW1wbCo+OjpIYXNoLCBIYXNoVHJhaXRzPEF0 b21pY1N0cmluZ0ltcGwqPiwgRW1wdHkxSW50SGFzaFRyYWl0cz4gVGFnUHJpb3JpdHlNYXA7CisK IGludCBIVE1MRWxlbWVudDo6dGFnUHJpb3JpdHkoKSBjb25zdAogewotICAgIGlmIChoYXNMb2Nh bE5hbWUod2JyVGFnKSkKLSAgICAgICAgcmV0dXJuIDA7Ci0gICAgaWYgKGhhc0xvY2FsTmFtZShh ZGRyZXNzVGFnKSB8fCBoYXNMb2NhbE5hbWUoZGRUYWcpIHx8IGhhc0xvY2FsTmFtZShkdFRhZykg fHwgaGFzTG9jYWxOYW1lKG5vc2NyaXB0VGFnKSB8fCBoYXNMb2NhbE5hbWUocnBUYWcpIHx8IGhh c0xvY2FsTmFtZShydFRhZykpCi0gICAgICAgIHJldHVybiAzOwotICAgIGlmIChoYXNMb2NhbE5h bWUoYXJ0aWNsZVRhZykgfHwgaGFzTG9jYWxOYW1lKGFzaWRlVGFnKSB8fCBoYXNMb2NhbE5hbWUo Y2VudGVyVGFnKSB8fCBoYXNMb2NhbE5hbWUoZm9vdGVyVGFnKSB8fCBoYXNMb2NhbE5hbWUoaGVh ZGVyVGFnKSB8fCBoYXNMb2NhbE5hbWUobm9iclRhZykgfHwgaGFzTG9jYWxOYW1lKHJ1YnlUYWcp IHx8IGhhc0xvY2FsTmFtZShuYXZUYWcpIHx8IGhhc0xvY2FsTmFtZShzZWN0aW9uVGFnKSkKLSAg ICAgICAgcmV0dXJuIDU7IC8vIFNhbWUgYXMgPGRpdj4uCi0gICAgaWYgKGhhc0xvY2FsTmFtZShu b2VtYmVkVGFnKSB8fCBoYXNMb2NhbE5hbWUobm9mcmFtZXNUYWcpKQotICAgICAgICByZXR1cm4g MTA7CisgICAgREVGSU5FX1NUQVRJQ19MT0NBTChUYWdQcmlvcml0eU1hcCwgdGFnUHJpb3JpdHlN YXAsICgpKTsKKyAgICBpZiAodGFnUHJpb3JpdHlNYXAuaXNFbXB0eSgpKSB7CisgICAgICAgIHRh Z1ByaW9yaXR5TWFwLmFkZCh3YnJUYWcubG9jYWxOYW1lKCkuaW1wbCgpLCAwKTsKKworICAgICAg ICB0YWdQcmlvcml0eU1hcC5hZGQoYWRkcmVzc1RhZy5sb2NhbE5hbWUoKS5pbXBsKCksIDMpOwor ICAgICAgICB0YWdQcmlvcml0eU1hcC5hZGQoZGRUYWcubG9jYWxOYW1lKCkuaW1wbCgpLCAzKTsK KyAgICAgICAgdGFnUHJpb3JpdHlNYXAuYWRkKGR0VGFnLmxvY2FsTmFtZSgpLmltcGwoKSwgMyk7 CisgICAgICAgIHRhZ1ByaW9yaXR5TWFwLmFkZChub3NjcmlwdFRhZy5sb2NhbE5hbWUoKS5pbXBs KCksIDMpOworICAgICAgICB0YWdQcmlvcml0eU1hcC5hZGQocnBUYWcubG9jYWxOYW1lKCkuaW1w bCgpLCAzKTsKKyAgICAgICAgdGFnUHJpb3JpdHlNYXAuYWRkKHJ0VGFnLmxvY2FsTmFtZSgpLmlt cGwoKSwgMyk7CisKKyAgICAgICAgLy8gNSBpcyBzYW1lIGFzIDxkaXY+J3MgcHJpb3JpdHkuCisg ICAgICAgIHRhZ1ByaW9yaXR5TWFwLmFkZChhcnRpY2xlVGFnLmxvY2FsTmFtZSgpLmltcGwoKSwg NSk7CisgICAgICAgIHRhZ1ByaW9yaXR5TWFwLmFkZChhc2lkZVRhZy5sb2NhbE5hbWUoKS5pbXBs KCksIDUpOworICAgICAgICB0YWdQcmlvcml0eU1hcC5hZGQoY2VudGVyVGFnLmxvY2FsTmFtZSgp LmltcGwoKSwgNSk7CisgICAgICAgIHRhZ1ByaW9yaXR5TWFwLmFkZChmb290ZXJUYWcubG9jYWxO YW1lKCkuaW1wbCgpLCA1KTsKKyAgICAgICAgdGFnUHJpb3JpdHlNYXAuYWRkKGhlYWRlclRhZy5s b2NhbE5hbWUoKS5pbXBsKCksIDUpOworICAgICAgICB0YWdQcmlvcml0eU1hcC5hZGQobm9iclRh Zy5sb2NhbE5hbWUoKS5pbXBsKCksIDUpOworICAgICAgICB0YWdQcmlvcml0eU1hcC5hZGQocnVi eVRhZy5sb2NhbE5hbWUoKS5pbXBsKCksIDUpOworICAgICAgICB0YWdQcmlvcml0eU1hcC5hZGQo bmF2VGFnLmxvY2FsTmFtZSgpLmltcGwoKSwgNSk7CisgICAgICAgIHRhZ1ByaW9yaXR5TWFwLmFk ZChzZWN0aW9uVGFnLmxvY2FsTmFtZSgpLmltcGwoKSwgNSk7CisKKyAgICAgICAgdGFnUHJpb3Jp dHlNYXAuYWRkKG5vZW1iZWRUYWcubG9jYWxOYW1lKCkuaW1wbCgpLCAxMCk7CisgICAgICAgIHRh Z1ByaW9yaXR5TWFwLmFkZChub2ZyYW1lc1RhZy5sb2NhbE5hbWUoKS5pbXBsKCksIDEwKTsKKwor ICAgICAgICAvLyBSZXR1cm4gMSBmb3Igb3RoZXIgdGFncy4gSXQncyBzYW1lIGFzIDxzcGFuPi4K KyAgICAgICAgLy8gVGhpcyB3YXkgY3VzdG9tIHRhZyBuYW1lIGVsZW1lbnRzIHdpbGwgYmVoYXZl IGxpa2UgaW5saW5lIHNwYW5zLgorICAgIH0KIAotICAgIC8vIFNhbWUgdmFsdWVzIGFzIDxzcGFu Pi4gIFRoaXMgd2F5IGN1c3RvbSB0YWcgbmFtZSBlbGVtZW50cyB3aWxsIGJlaGF2ZSBsaWtlIGlu bGluZSBzcGFucy4KLSAgICByZXR1cm4gMTsKKyAgICByZXR1cm4gdGFnUHJpb3JpdHlNYXAuZ2V0 KGxvY2FsTmFtZSgpLmltcGwoKSk7CiB9CiAKIGJvb2wgSFRNTEVsZW1lbnQ6Om1hcFRvRW50cnko Y29uc3QgUXVhbGlmaWVkTmFtZSYgYXR0ck5hbWUsIE1hcHBlZEF0dHJpYnV0ZUVudHJ5JiByZXN1 bHQpIGNvbnN0Ci0tIAoxLjYuMy4zCgo= 45999 2010-01-06 15:02:36 -0800 2010-01-06 16:26:36 -0800 Proposed patch (rev.2) 0001-element-hash-2.patch text/plain 4074 tkent RnJvbSA5Njc2MTYzOGMzMWM1NGNmY2JjNjE3NWFhOTYyOWU0Njg1N2MxNzdhIE1vbiBTZXAgMTcg MDA6MDA6MDAgMjAwMQpGcm9tOiBLZW50IFRhbXVyYSA8dGtlbnRAY2hyb21pdW0ub3JnPgpEYXRl OiBUaHUsIDcgSmFuIDIwMTAgMDY6NDk6NTYgKzA5MDAKU3ViamVjdDogW1BBVENIXSBlbGVtZW50 LWhhc2gtMgoKLS0tCiBXZWJDb3JlL0NoYW5nZUxvZyAgICAgICAgICAgIHwgICAxNyArKysrKysr KysrKysrKwogV2ViQ29yZS9odG1sL0hUTUxFbGVtZW50LmNwcCB8ICAgNTEgKysrKysrKysrKysr KysrKysrKysrKysrKysrKysrKystLS0tLS0tLS0KIDIgZmlsZXMgY2hhbmdlZCwgNTcgaW5zZXJ0 aW9ucygrKSwgMTEgZGVsZXRpb25zKC0pCgpkaWZmIC0tZ2l0IGEvV2ViQ29yZS9DaGFuZ2VMb2cg Yi9XZWJDb3JlL0NoYW5nZUxvZwppbmRleCAyNTViZDU4Li4xZmJlNjJiIDEwMDY0NAotLS0gYS9X ZWJDb3JlL0NoYW5nZUxvZworKysgYi9XZWJDb3JlL0NoYW5nZUxvZwpAQCAtMSwzICsxLDIwIEBA CisyMDEwLTAxLTA2ICBLZW50IFRhbXVyYSAgPHRrZW50QGNocm9taXVtLm9yZz4KKworICAgICAg ICBSZXZpZXdlZCBieSBOT0JPRFkgKE9PUFMhKS4KKworICAgICAgICBVc2UgYSBzdGF0aWMgSGFz aE1hcCBmb3IgSFRNTEVsZW1lbnQ6OnRhZ1ByaW9yaXR5KCkuCisgICAgICAgIGh0dHBzOi8vYnVn cy53ZWJraXQub3JnL3Nob3dfYnVnLmNnaT9pZD0zMzI2OQorCisgICAgICAgIFRoZSBwcmlvciBj b2RlIGNvbXBhcmVzIEF0b21pY1N0cmluZ0ltcGwgcG9pbnRlcnMgMTggdGltZXMgYXQKKyAgICAg ICAgd29yc3QuIFRoaXMgY2hhbmdlIGF2b2lkcyBpdC4KKworICAgICAgICBObyBuZXcgdGVzdHMg YmVjYXVzZSB0aGlzIGlzIGp1c3QgYSByZWZhY3RvcmluZy4KKworICAgICAgICAqIGh0bWwvSFRN TEVsZW1lbnQuY3BwOgorICAgICAgICAoV2ViQ29yZTo6RW1wdHkxSW50SGFzaFRyYWl0cyk6IEEg SGFzaFRyYWl0cyB0byByZXR1cm4gMSBhcyB0aGUgZW1wdHkgdmFsdWUuCisgICAgICAgIChXZWJD b3JlOjppbml0aWFsaXplVGFnUHJpb3JpdHlNYXApOiBJbml0aWFsaXphdGlvbiBvZiBhIHN0YXRp YyBIYXNoTWFwLgorICAgICAgICAoV2ViQ29yZTo6SFRNTEVsZW1lbnQ6OnRhZ1ByaW9yaXR5KTog VXNlIHRoZSBzdGF0aWMgSGFzaE1hcCBjcmVhdGVkIGJ5IGluaXRpYWxpemVUYWdQcmlvcml0eU1h cCgpLgorCiAyMDEwLTAxLTA2ICBTaW1vbiBIYXVzbWFubiAgPHNpbW9uLmhhdXNtYW5uQG5va2lh LmNvbT4KIAogICAgICAgICBVbnJldmlld2VkIHRyaXZpYWwgUXQgYnVpbGQgZml4LgpkaWZmIC0t Z2l0IGEvV2ViQ29yZS9odG1sL0hUTUxFbGVtZW50LmNwcCBiL1dlYkNvcmUvaHRtbC9IVE1MRWxl bWVudC5jcHAKaW5kZXggZmZhMGVlYS4uZjVhNWRmOSAxMDA2NDQKLS0tIGEvV2ViQ29yZS9odG1s L0hUTUxFbGVtZW50LmNwcAorKysgYi9XZWJDb3JlL2h0bWwvSFRNTEVsZW1lbnQuY3BwCkBAIC04 MiwxOSArODIsNDggQEAgSFRNTFRhZ1N0YXR1cyBIVE1MRWxlbWVudDo6ZW5kVGFnUmVxdWlyZW1l bnQoKSBjb25zdAogICAgIHJldHVybiBUYWdTdGF0dXNSZXF1aXJlZDsKIH0KIAotaW50IEhUTUxF bGVtZW50Ojp0YWdQcmlvcml0eSgpIGNvbnN0CitzdHJ1Y3QgRW1wdHkxSW50SGFzaFRyYWl0cyA6 IEhhc2hUcmFpdHM8aW50PiB7CisgICAgc3RhdGljIGNvbnN0IGJvb2wgZW1wdHlWYWx1ZUlzWmVy byA9IGZhbHNlOworICAgIHN0YXRpYyBpbnQgZW1wdHlWYWx1ZSgpIHsgcmV0dXJuIDE7IH0KK307 CisvLyBUYWdQcmlvcml0eU1hcCByZXR1cm5zIDEgZm9yIHVucmVnaXN0ZXJlZCB0YWdzLiBJdCdz IHNhbWUgYXMgPHNwYW4+LgorLy8gVGhpcyB3YXkgY3VzdG9tIHRhZyBuYW1lIGVsZW1lbnRzIHdp bGwgYmVoYXZlIGxpa2UgaW5saW5lIHNwYW5zLgordHlwZWRlZiBIYXNoTWFwPEF0b21pY1N0cmlu Z0ltcGwqLCBpbnQsIERlZmF1bHRIYXNoPEF0b21pY1N0cmluZ0ltcGwqPjo6SGFzaCwgSGFzaFRy YWl0czxBdG9taWNTdHJpbmdJbXBsKj4sIEVtcHR5MUludEhhc2hUcmFpdHM+IFRhZ1ByaW9yaXR5 TWFwOworCitzdGF0aWMgY29uc3QgVGFnUHJpb3JpdHlNYXAqIGluaXRpYWxpemVUYWdQcmlvcml0 eU1hcCgpCiB7Ci0gICAgaWYgKGhhc0xvY2FsTmFtZSh3YnJUYWcpKQotICAgICAgICByZXR1cm4g MDsKLSAgICBpZiAoaGFzTG9jYWxOYW1lKGFkZHJlc3NUYWcpIHx8IGhhc0xvY2FsTmFtZShkZFRh ZykgfHwgaGFzTG9jYWxOYW1lKGR0VGFnKSB8fCBoYXNMb2NhbE5hbWUobm9zY3JpcHRUYWcpIHx8 IGhhc0xvY2FsTmFtZShycFRhZykgfHwgaGFzTG9jYWxOYW1lKHJ0VGFnKSkKLSAgICAgICAgcmV0 dXJuIDM7Ci0gICAgaWYgKGhhc0xvY2FsTmFtZShhcnRpY2xlVGFnKSB8fCBoYXNMb2NhbE5hbWUo YXNpZGVUYWcpIHx8IGhhc0xvY2FsTmFtZShjZW50ZXJUYWcpIHx8IGhhc0xvY2FsTmFtZShmb290 ZXJUYWcpIHx8IGhhc0xvY2FsTmFtZShoZWFkZXJUYWcpIHx8IGhhc0xvY2FsTmFtZShub2JyVGFn KSB8fCBoYXNMb2NhbE5hbWUocnVieVRhZykgfHwgaGFzTG9jYWxOYW1lKG5hdlRhZykgfHwgaGFz TG9jYWxOYW1lKHNlY3Rpb25UYWcpKQotICAgICAgICByZXR1cm4gNTsgLy8gU2FtZSBhcyA8ZGl2 Pi4KLSAgICBpZiAoaGFzTG9jYWxOYW1lKG5vZW1iZWRUYWcpIHx8IGhhc0xvY2FsTmFtZShub2Zy YW1lc1RhZykpCi0gICAgICAgIHJldHVybiAxMDsKKyAgICBUYWdQcmlvcml0eU1hcCogbWFwID0g bmV3IFRhZ1ByaW9yaXR5TWFwOworCisgICAgbWFwLT5hZGQod2JyVGFnLmxvY2FsTmFtZSgpLmlt cGwoKSwgMCk7CisKKyAgICBtYXAtPmFkZChhZGRyZXNzVGFnLmxvY2FsTmFtZSgpLmltcGwoKSwg Myk7CisgICAgbWFwLT5hZGQoZGRUYWcubG9jYWxOYW1lKCkuaW1wbCgpLCAzKTsKKyAgICBtYXAt PmFkZChkdFRhZy5sb2NhbE5hbWUoKS5pbXBsKCksIDMpOworICAgIG1hcC0+YWRkKG5vc2NyaXB0 VGFnLmxvY2FsTmFtZSgpLmltcGwoKSwgMyk7CisgICAgbWFwLT5hZGQocnBUYWcubG9jYWxOYW1l KCkuaW1wbCgpLCAzKTsKKyAgICBtYXAtPmFkZChydFRhZy5sb2NhbE5hbWUoKS5pbXBsKCksIDMp OworCisgICAgLy8gNSBpcyBzYW1lIGFzIDxkaXY+J3MgcHJpb3JpdHkuCisgICAgbWFwLT5hZGQo YXJ0aWNsZVRhZy5sb2NhbE5hbWUoKS5pbXBsKCksIDUpOworICAgIG1hcC0+YWRkKGFzaWRlVGFn LmxvY2FsTmFtZSgpLmltcGwoKSwgNSk7CisgICAgbWFwLT5hZGQoY2VudGVyVGFnLmxvY2FsTmFt ZSgpLmltcGwoKSwgNSk7CisgICAgbWFwLT5hZGQoZm9vdGVyVGFnLmxvY2FsTmFtZSgpLmltcGwo KSwgNSk7CisgICAgbWFwLT5hZGQoaGVhZGVyVGFnLmxvY2FsTmFtZSgpLmltcGwoKSwgNSk7Cisg ICAgbWFwLT5hZGQobm9iclRhZy5sb2NhbE5hbWUoKS5pbXBsKCksIDUpOworICAgIG1hcC0+YWRk KHJ1YnlUYWcubG9jYWxOYW1lKCkuaW1wbCgpLCA1KTsKKyAgICBtYXAtPmFkZChuYXZUYWcubG9j YWxOYW1lKCkuaW1wbCgpLCA1KTsKKyAgICBtYXAtPmFkZChzZWN0aW9uVGFnLmxvY2FsTmFtZSgp LmltcGwoKSwgNSk7CisKKyAgICBtYXAtPmFkZChub2VtYmVkVGFnLmxvY2FsTmFtZSgpLmltcGwo KSwgMTApOworICAgIG1hcC0+YWRkKG5vZnJhbWVzVGFnLmxvY2FsTmFtZSgpLmltcGwoKSwgMTAp OworCisgICAgcmV0dXJuIG1hcDsKK30KIAotICAgIC8vIFNhbWUgdmFsdWVzIGFzIDxzcGFuPi4g IFRoaXMgd2F5IGN1c3RvbSB0YWcgbmFtZSBlbGVtZW50cyB3aWxsIGJlaGF2ZSBsaWtlIGlubGlu ZSBzcGFucy4KLSAgICByZXR1cm4gMTsKK2ludCBIVE1MRWxlbWVudDo6dGFnUHJpb3JpdHkoKSBj b25zdAoreworICAgIHN0YXRpYyBjb25zdCBUYWdQcmlvcml0eU1hcCogdGFnUHJpb3JpdHlNYXAg PSBpbml0aWFsaXplVGFnUHJpb3JpdHlNYXAoKTsKKyAgICByZXR1cm4gdGFnUHJpb3JpdHlNYXAt PmdldChsb2NhbE5hbWUoKS5pbXBsKCkpOwogfQogCiBib29sIEhUTUxFbGVtZW50OjptYXBUb0Vu dHJ5KGNvbnN0IFF1YWxpZmllZE5hbWUmIGF0dHJOYW1lLCBNYXBwZWRBdHRyaWJ1dGVFbnRyeSYg cmVzdWx0KSBjb25zdAotLSAKMS42LjMuMwoK