User:SadEagle: Difference between revisions

    From KDE TechBase
     
    (75 intermediate revisions by 2 users not shown)
    Line 1: Line 1:
    == In-flight stuff ==
    * from olliej:
    PropertySlot::getValue(ExecState, unsigned propertyName) converts propertyName to an Identifier, despite the whole reason for that separate getValue being to avoid int->identifier conversion
    [03:40] <olliej> SadEagle: if you add the concept of an indexing getter function you can avoid that conversion cost (it's significant)
    * <b> Consolidate my other notes</b>
    * Fix innerHTML serialization of script and similar.
    ** also in comments
    * show/hide in KHTMLPart
    * <s>relaxing checks in innerHTML, for google.com</s>
    ** <s>but need to debug follow up problem with clicks</s>
    *** <s>seems to be fixed independently. </s>
    ** Need testcase
    * namespace in cloneNode, for the SVG use assert failure
    * click/KHTML_ECMA_CLICK --- and doubleclick cleanup (bug report by pino's brother on openstreetmaps)
    * DOMTS cleanup
    * <s>combobox sizing -- making the popup expand.</s>
    * <s>regtest carewolf's lt patch.</s>
    * scroll, and in general window events (slashdot autoexpand)
    ** something more there, too.
    * Java/security exception.
    * caught exception in the JS debugger
    * ID generation? (ade's stuff, xlink namespace)
    * Debug the open-profile/statusbar mess.
    * sidebar dom tree viewer?
    * event handlers not logging exceptions
    * <s>string interning.</s>
    * <s>split domTreeVersion to have structural and specific attribute versions;
    make use of that to make ClassNodeList cacheable.</s>
    * <s>implement compareDocumentPosition</s>
    == Next in queue ==
    * Performance issues with certificate loading
    * computeContent family of crashes
    * look at the readonly/focus thing again
    * Log various load events, profile stuff.
    * <s>If spart doesn't get to it, / and related stuff --- biggest
    stability issue</s>
    * Fix SVG viewport thing
    * <s>Bero had a patch to btoa.cpp</s>
    * Figure out memory management for XHR.
    * new autogen
    ** should help with XPath stuff.
    ** don't forget bindings stuff
    * multitab thing
    * on google groups: key events with target element being the document.
    == Current priorities ==
    == Current priorities ==
    * [[Get]] for NamedNodeMap
    * new request methods OK'd for XHR
    * For 4.2, do bindings keys, notify kde-bindings/rdale
    * Commit of {{Bug|150843}} fix is long overdue
    * Other high-priority bugs:
    ** The derstandard.at one
    ** In general recursion of tokenizer, scripts. e.g. {{Bug|170765}}
    ** Figure out the selector crash Frank R. testcased, too {{Bug|150662}}
    * head/body parsing
    * Fix namespacing of WTF
    * Fix namespacing of WTF
    * Backport the comment lexing fix
    * Look at google groups again
    * <s>Do the nspluginviewer exit fix</s>
    * I also had mem-use reduction change for KJS, which had the issue of the global code's lack of body for rewinding.  Do we ever run code in GlobalExec, though? We shouldn't...
    * Finish up the FrameArray removal, as below and also for the crasher
    * There was also array iteration stuff, not sure it matters, though
    * Disabled color for Roman
    * history/:visited and acid3
    * Accel handling?
    * Fix const stuff
    * Pick up the getComputedStyle stuff from WC
    * <s>The color inheritance stuff... Also, makes me wonder about currentColor --- that's tied in to gcs, though.</s>
    * the adblock stuff is in progress, but probably should be trunk-only.
    ** Spart fixed that and added currentColor parsing in trunk. So with GCS I should be able to add it to canvas.
    * I also had mem-use reduction change for KJS, which had the issue of  
    * KMail drag icon: http://lxr.kde.org/source/KDE/kdepim/kmail/kmheaders.cpp#2419
    the global code's lack of body for rewinding.  Do we ever run  
    * Something is wrong with abcnews.com
    code in GlobalExec, though? We shouldn't
     
    == Performance stuff ==
    * Qt X509 parsing/CA bundle loading
    * setClipRect inefficiency --- report to TT
    * Make khtml use setClipRect less.


    == Misc stuff==
    == Misc stuff==
    Line 18: Line 77:


    == nspluginviewer ==
    == nspluginviewer ==
    * Debug recent regressions
    * Debug recent regressions... if I could reproduce it
    * <s>FP flash 10 workaround</s>
    * Try to catch Seli about the SuSE/GlibEvents path. Or sic the reporters on the SuSE bugzilla.
    * <s>Fix exiting on host exit, double-check some more  to make sure no process leaks
    ** serviceOwnerChanged
    ** dfaure:  See Kded::slotApplicationRemoved for an example;  (the possibly unintuitive bit is "if (oldOwner.isEmpty() || !newOwner.isEmpty()) return;"</s>


    == Canvas ==
    == Canvas ==
    Line 28: Line 84:
    * Update error handling?
    * Update error handling?
    * setFillColor(double r, double g, double b, double a) extension
    * setFillColor(double r, double g, double b, double a) extension
    * Can now make width/height reflecting, with ::rewriteValue


    == KJS and KJS/Frostbyte ==
    == KJS and KJS/Frostbyte ==
    * Idea: don't use hashtables for prototype objects' function fields. This way they can benefit from inline caching
    ** Uhm, how do we manage the prototypechain lookup? Seems easier with a type system
    * Idea: function pointers in propertyMap? Makes the map fat, but then can unifty... what? forgot already. Well, certainly getters... Hmm, may be dynamic properties can be handled in the prototype like that, getter-like?
    * Idea: split off the hash table stuff off from JSObject
    * Idea: split off the hash table stuff off from JSObject
    ** Probably not workable since the prototype even of pure objects  can have hashtable properties. Object.prototype doesn't, however, so  we can bypass a lot of stuff. Properties there can only be DontEnum or default, too.  
    ** Probably not workable since the prototype even of pure objects  can have hashtable properties. Object.prototype doesn't, however, so  we can bypass a lot of stuff. Properties there can only be DontEnum or default, too.  
    * Try to do inline-in-ExecState scopeChainNodes again --- but can they be torn off?
    ** May be store in the action? Seems illegal type-wise.
    * Debugger, again
    * Debugger, again
    ** Mostly works, but stepping through ifs is funny --- need to figure out why
    ** Mostly works, but stepping through ifs is funny --- need to figure out why
    Line 71: Line 129:


    == KHTML - Short-term stuff ==
    == KHTML - Short-term stuff ==
    * Don't forget the clip:rect(auto, 123, ...) stuff
    * Don't forget the leak fixes
    * Don't forget the leak fixes
    ** consult with Maciej and Micha
    ** consult with Maciej and Micha
    Line 79: Line 136:
    * Fix the damn image-load-event-thing in iframes
    * Fix the damn image-load-event-thing in iframes
    * Go through the canvas testcase again
    * Go through the canvas testcase again
    * <s>Fix noscript on DOM ops. Should it just display:none? I guess I should check HTML5</s>
    * Important, don't know how to fix: {{Bug|163359}}
    * Important, don't know how to fix: {{Bug|163359}}
    * Need to debug various editors
    * {{Bug|162745}} -- for CSS, we need to strip whitespace from URLs. Ditto for object, I'd think. Perhaps completeURL should do it.  
    * {{Bug|162745}} -- for CSS, we need to strip whitespace from URLs. Ditto for object, I'd think. Perhaps completeURL should do it.  
    ** {{Bug|161587}} -- for images, too!
    ** {{Bug|161587}} -- for images, too!
    * For objects, empty URLs should probably embed still -- but when? CompleteURL on empty seems like a culprit for the self-embed bugs.  
    * For objects, empty URLs should probably embed still -- but when? CompleteURL on empty seems like a culprit for the self-embed bugs.  
    * Inherit params  
    * Inherit params  
    * <s>{{Bug|161196}} --- option value --- should be super-quick</s>
    * JS Redirects
    * JS Redirects
    ** <s>Anchor-go canceling parser {{Bug|159192}} --- Wikipedia!</s>
    ** <s>More synchrony/timing {{Bug|159279}} </s>
    ** Ordering with respect to form submissions {{Bug|159932}}
    ** Ordering with respect to form submissions {{Bug|159932}}
    ** Ordering with respect to http {{Bug|132031}}
    ** Ordering with respect to http {{Bug|132031}}
    ** WRT to link clicks {{Bug|120157}}
    ** WRT to link clicks {{Bug|120157}}
    ** Check: {{Bug|109694}} --- likely fixed
    ** Check: {{Bug|109694}} --- likely fixed
    ** <s>window.location value: {{Bug|137176}}</s>
    ** {{Bug|112756}}: affects the wrong part?
    ** {{Bug|112756}}: affects the wrong part?
    * Iframe stuff
    * Iframe stuff
    Line 107: Line 158:
    ** {{Bug|158036}}?
    ** {{Bug|158036}}?
    ** Check: {{Bug|116189}}?
    ** Check: {{Bug|116189}}?
    * Fix the paranoia affecting(?) flickr(?) upload {{Bug|126974}}
    * The XHTML parser mode
    * The XHTML parser mode
    ** {{Bug|164867}}
    ** {{Bug|164867}}
    Line 126: Line 176:


    == KHTML - Long term stuff ==
    == KHTML - Long term stuff ==
    * Fix adblock performance. At the very least, special case substring matching to not use QRegExp.
    * Consider changing window.frames to be == to window
    ** {{Bug|168017}}
    ** {{Bug|77966}}?
    * CSS Loading stuff
    * CSS Loading stuff
    ** Spart fixed a bunch of stuff already
    ** Spart fixed a bunch of stuff already
    Line 152: Line 198:
    == Other kdelibs stuff ==
    == Other kdelibs stuff ==
    * Figure out why the maxProcess stuff in kio_http doesn't seem to work.
    * Figure out why the maxProcess stuff in kio_http doesn't seem to work.
    == KJS game plan ==
    * Implement eval semantics from ES5 (presumably)
    * Make sure to track down the patch that adds $hint$, lookup stuff, maybe even paste it here to have a backup
    * Scoping issues:
    ** think (formalize?) function scope stuff
    ** alterations for function literals, catch scope stuff in
    ** rework how we handle declarations to avoid visitors?
    ** blizzard stage 2 stuff...
    ** public hook to cache document?
    * Function call stuff:
    ** Get rid of JSObject::call first; or at least move the depth tracking out of it and into Machine (easy, independent)
    ** olliej says to push parameters in reverse... think of it some more.
    ** look at slimming down ExecState again...
    ** change how List works to permit its usage of JS stack.
    ** remember the dynamic link/static link distinction: the scope chain is the static link, and can we disconnected, but the dynamic link will always be to parent frame
    *** sort of. If one doesn't do a native JS call, one may have to copy stuff over. So essentially it's a 2 case thing: either our stack frame is immediately after, with arguments on the parent frame, or when we do a call from KHTML, etc., we also copy over arguments... in which case, well... i guess we would need two pointers to the frame or something --- one for marking, and one for giving the interpreter?

    Latest revision as of 16:04, 2 March 2010

    In-flight stuff

    • from olliej:
    PropertySlot::getValue(ExecState, unsigned propertyName) converts propertyName to an Identifier, despite the whole reason for that separate getValue being to avoid int->identifier conversion
    

    [03:40] <olliej> SadEagle: if you add the concept of an indexing getter function you can avoid that conversion cost (it's significant)

    • Consolidate my other notes
    • Fix innerHTML serialization of script and similar.
      • also in comments
    • show/hide in KHTMLPart
    • relaxing checks in innerHTML, for google.com
      • but need to debug follow up problem with clicks
        • seems to be fixed independently.
      • Need testcase
    • namespace in cloneNode, for the SVG use assert failure
    • click/KHTML_ECMA_CLICK --- and doubleclick cleanup (bug report by pino's brother on openstreetmaps)
    • DOMTS cleanup
    • combobox sizing -- making the popup expand.
    • regtest carewolf's lt patch.
    • scroll, and in general window events (slashdot autoexpand)
      • something more there, too.
    • Java/security exception.
    • caught exception in the JS debugger
    • ID generation? (ade's stuff, xlink namespace)
    • Debug the open-profile/statusbar mess.
    • sidebar dom tree viewer?
    • event handlers not logging exceptions
    • string interning.
    • split domTreeVersion to have structural and specific attribute versions;

    make use of that to make ClassNodeList cacheable.

    • implement compareDocumentPosition

    Next in queue

    • Performance issues with certificate loading
    • computeContent family of crashes
    • look at the readonly/focus thing again
    • Log various load events, profile stuff.
    • If spart doesn't get to it, / and related stuff --- biggest

    stability issue

    • Fix SVG viewport thing
    • Bero had a patch to btoa.cpp
    • Figure out memory management for XHR.
    • new autogen
      • should help with XPath stuff.
      • don't forget bindings stuff
    • multitab thing
    • on google groups: key events with target element being the document.

    Current priorities

    • Get for NamedNodeMap
    • new request methods OK'd for XHR
    • For 4.2, do bindings keys, notify kde-bindings/rdale
    • Commit of bug #150843 fix is long overdue
    • Other high-priority bugs:
      • The derstandard.at one
      • In general recursion of tokenizer, scripts. e.g. bug #170765
      • Figure out the selector crash Frank R. testcased, too bug #150662
    • head/body parsing
    • Fix namespacing of WTF
    • Look at google groups again
    • I also had mem-use reduction change for KJS, which had the issue of the global code's lack of body for rewinding. Do we ever run code in GlobalExec, though? We shouldn't...
    • There was also array iteration stuff, not sure it matters, though
    • history/:visited and acid3
    • Fix const stuff
    • The color inheritance stuff... Also, makes me wonder about currentColor --- that's tied in to gcs, though.
      • Spart fixed that and added currentColor parsing in trunk. So with GCS I should be able to add it to canvas.
    • KMail drag icon: http://lxr.kde.org/source/KDE/kdepim/kmail/kmheaders.cpp#2419
    • Something is wrong with abcnews.com

    Performance stuff

    • Qt X509 parsing/CA bundle loading
    • setClipRect inefficiency --- report to TT
    • Make khtml use setClipRect less.

    Misc stuff

    nspluginviewer

    • Debug recent regressions... if I could reproduce it
    • Try to catch Seli about the SuSE/GlibEvents path. Or sic the reporters on the SuSE bugzilla.

    Canvas

    • Update for createImageData, range changes
    • Update error handling?
    • setFillColor(double r, double g, double b, double a) extension
    • Can now make width/height reflecting, with ::rewriteValue

    KJS and KJS/Frostbyte

    • Idea: don't use hashtables for prototype objects' function fields. This way they can benefit from inline caching
      • Uhm, how do we manage the prototypechain lookup? Seems easier with a type system
    • Idea: function pointers in propertyMap? Makes the map fat, but then can unifty... what? forgot already. Well, certainly getters... Hmm, may be dynamic properties can be handled in the prototype like that, getter-like?
    • Idea: split off the hash table stuff off from JSObject
      • Probably not workable since the prototype even of pure objects can have hashtable properties. Object.prototype doesn't, however, so we can bypass a lot of stuff. Properties there can only be DontEnum or default, too.
    • Debugger, again
      • Mostly works, but stepping through ifs is funny --- need to figure out why
    • Unbreak CPU guard.
    • See if we can keep track of the current function body in the parser. Use that to help track of things like presence of "arguments"
      • With this and eval operator, may be able to get rid of an explicit activation and scope chain entry...
    • Type system changes:
      • Make int32 immediates apply to int32 values --- add a notion of exact embedding
      • Add a notion of value/number dual-use in registers --- use DontMark bit to distinguish which it is
    • Idea: have a template for activation in FunctionImp, and then just memcpy, minus the params, on startup
      • still need to setup functions separately, though.
      • seems slower, actually
    • Idea, part 2: use the above to setup immediates in the same place as registers. This can simplify the IR a lot and shrink the loop considerably. Also means there is no longer architecture differences in 32-bit and 64-bit needs at the IR level, and the IR gets smaller.
      • Of course, recursion more expensive. But worth a try?
    • BUG: need to flush lazy local copies at function call (or do I do it already?)
    • Alternative/conflict with below: if using a single stack for parameter passing/direct calls, we want to always do stack allocation and then do a deferred tearoff when needed --- this may actually be visible based on refcount of the scope chain, except when manually done.
      • That's actually done, but I am not sure of how to best combine it with other stuff
    • Do segmented/nonmoveable stack allocation --- just have a list of pieces, and work on them one at a time. Older ones can be recycled. This means we never have to refetch the locals array.Also avoids the WARNING WARNING WARNING part.
    • Do deferred tearoffs.
    • eval operator/eval tearoffs
      • tearoffs done.
      • With eval operator, if I keep track of lexical depth, may be able to directly route to global object, etc. Needs thought.
    • Fix const
    • obscure sequencing thing
      • seems like an another regression on the TC though --- exception precision?
    • merge in the API, etc., from trunk
    • Do the List performance changes
    • Do IR dialectes for 32- and 64-bit -- e.g. PointerCell
      • Also probably double-allocator for 32-bit.
    • Do a review pass over API to make sure not BC leaks.
    • Double-check global handling, synchronize with JSC
    • Consider using packed immediate/register bits. The key observation is that locals are read from an array, too, so we could do something like:
      arg0 = argsType[(fullOp >> 16) & 1][offsetVal];
    
    • Crazy idea: when going to proper CFG construction, do we even have to linearize?

    KHTML - Short-term stuff

    • Don't forget the leak fixes
      • consult with Maciej and Micha
    • There was a change in getElementsByClassName case behavior
    • Do floating-point font size computation
      • Or rather e-mail it to kfm-devel
    • Fix the damn image-load-event-thing in iframes
    • Go through the canvas testcase again
    • Important, don't know how to fix: bug #163359
    • bug #162745 -- for CSS, we need to strip whitespace from URLs. Ditto for object, I'd think. Perhaps completeURL should do it.
    • For objects, empty URLs should probably embed still -- but when? CompleteURL on empty seems like a culprit for the self-embed bugs.
    • Inherit params
    • JS Redirects
    • Iframe stuff
    • The XHTML parser mode
    • Check the XML parser -- QXML workaround might not be needed, weird whitespace handling and recovery stuff
    • Check the implicit node stuff. Only the form one seems correct to me, and that can be done w/o wasting a bit at top-level.
    • the libthai thing
    • bug #104358
    • XHR:
    • bug #103250 --- KWallet and pages with textareas.

    khtmlImLoad

    • Allan's foreign loader patch
    • Scale animation provider for GIF --- bug #162614
    • Animation stoping
    • Better pre-blend cache?
    • FredrikH performance goodness?

    KHTML - Long term stuff

    Other long term stuff

    • Async loading of nspluginviewer somehow?

    Other kdelibs stuff

    • Figure out why the maxProcess stuff in kio_http doesn't seem to work.

    KJS game plan

    • Implement eval semantics from ES5 (presumably)
    • Make sure to track down the patch that adds $hint$, lookup stuff, maybe even paste it here to have a backup
    • Scoping issues:
      • think (formalize?) function scope stuff
      • alterations for function literals, catch scope stuff in
      • rework how we handle declarations to avoid visitors?
      • blizzard stage 2 stuff...
      • public hook to cache document?
    • Function call stuff:
      • Get rid of JSObject::call first; or at least move the depth tracking out of it and into Machine (easy, independent)
      • olliej says to push parameters in reverse... think of it some more.
      • look at slimming down ExecState again...
      • change how List works to permit its usage of JS stack.
      • remember the dynamic link/static link distinction: the scope chain is the static link, and can we disconnected, but the dynamic link will always be to parent frame
        • sort of. If one doesn't do a native JS call, one may have to copy stuff over. So essentially it's a 2 case thing: either our stack frame is immediately after, with arguments on the parent frame, or when we do a call from KHTML, etc., we also copy over arguments... in which case, well... i guess we would need two pointers to the frame or something --- one for marking, and one for giving the interpreter?