summaryrefslogtreecommitdiff
path: root/desktop/inc/lib
AgeCommit message (Collapse)Author
2023-04-05fix leaks when using tools::JsonWriterNoel Grandin
Specifically in sd/source/core/annotations/Annotation.cxx We seem to end up fixing leaks here often. The current tools::JsonWriter API is just very hard to use correctly. So rather return an OString, which is cheap to copy, and push that down into the LOK code. AFAIK that seems to end up requiring less code and less adhoc copying of data (specifically the queueing code in init.cxx was creating copies when converting to std::string). Ideally, we could have some special API to avoid the new strdup() calls in init.cxx, but not sure how to prevent other people from accidentally using that. Change-Id: Ia33437c1bfd9cc2d54dfb99914d1b72db20335f2 Reviewed-on: https://gerrit.libreoffice.org/c/core/+/149963 Tested-by: Jenkins Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
2022-11-10lok: masterpage: use EditMode as parameter in invalidation callbackSzymon Kłos
Change-Id: I3905fc9e6376ca1cef3e438e7a5f229d3720b1f5 Reviewed-on: https://gerrit.libreoffice.org/c/core/+/138961 Tested-by: Jenkins CollaboraOffice <jenkinscollaboraoffice@gmail.com> Reviewed-by: Henry Castro <hcastro@collabora.com> Reviewed-on: https://gerrit.libreoffice.org/c/core/+/142501 Tested-by: Jenkins Reviewed-by: Szymon Kłos <szymon.klos@collabora.com>
2022-08-25Tell LibreOfficeKit clients what fonts in a document are missing on the machineTor Lillqvist
Use the OutputDevice::StartTrackingFontMappingUse() and OutputDevice::FinishTrackingFontMappingUse() functionality that was added last year. Add a new LibreOfficeKit document callback: LOK_CALLBACK_FONTS_MISSING. "Font" here means just the family name. We dont really know what style from the family is missing. Change-Id: Iec4349b5827e7d1dbdde7f8a9f8da92efde429e0 Reviewed-on: https://gerrit.libreoffice.org/c/core/+/138802 Tested-by: Jenkins Reviewed-by: Tor Lillqvist <tml@collabora.com>
2022-08-25lok: make sure flushPendingLOKInvalidateTiles() is calledLuboš Luňák
SwViewShellImp::AddPendingLOKInvalidation() collects invalidations, but this was relying on something eventually calling flushPendingLOKInvalidateTiles(), which wasn't guaranteed. If e.g. a spellchecker caused an invalidation from in idle callback and nothing else changed, then the LOK callback handling code didn't know there was something pending. So add an explicit call to ensure to notify about these pending invalidations. Change-Id: I0a9cb0d5aba2fdbbac126cd8a4a3412bef1cab25 Reviewed-on: https://gerrit.libreoffice.org/c/core/+/136531 Tested-by: Jenkins Reviewed-by: Luboš Luňák <l.lunak@collabora.com>
2022-07-28lok: add dumpState feature for better in-field diagnostics.Michael Meeks
Always suspicious that some un-expected dialog / state can cause strange behavior in a client. An initial cut at an API to make it easier to unwind such problems by exposing the toolkit state. Change-Id: If8f17943fa4837df4f9ca659a111dcdce5c23244 Reviewed-on: https://gerrit.libreoffice.org/c/core/+/137504 Tested-by: Jenkins Reviewed-by: Michael Meeks <michael.meeks@collabora.com>
2022-06-08use recursive mutex for LOK queueLuboš Luňák
Callbacks may be invoked while calling getLOKPayload(), which would try to lock the mutex again. I actually originally expected this possibility, as the comment and moving the data to temporaries in CallbackFlushHandler::enqueueUpdatedTypes() shows, I just didn't realize the used mutex wasn't recursive and so would deadlock. Change-Id: I2b5c4b6b4c1a3933a32ae4641830877e085f2b6c Reviewed-on: https://gerrit.libreoffice.org/c/core/+/135499 Tested-by: Jenkins Reviewed-by: Michael Meeks <michael.meeks@collabora.com>
2022-06-03clang-tidy modernize-pass-by-value in desktopNoel Grandin
Change-Id: Ie9c1cd41c5944b901f4ec6c1e4ffd8608c063ffa Reviewed-on: https://gerrit.libreoffice.org/c/core/+/135317 Tested-by: Jenkins Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
2022-02-25make sure that LOK "updated types" start a timer as wellLuboš Luňák
273a25c796fca9afa0dfadac57dc3f336831221c changed the handling of LOK messages to only set up an "updated" flag for types and then when flushing the value is fetched. This means that queue() is not called for such updates, and so the timeout to flush the messages wasn't started. Change-Id: I3a7324c86a55697f509f80b6c54cb949fb1daa72 Reviewed-on: https://gerrit.libreoffice.org/c/core/+/130526 Tested-by: Jenkins Reviewed-by: Luboš Luňák <l.lunak@collabora.com>
2022-01-20lok: avoid duplicate emission of statechanged: messages.Michael Meeks
We tend to get many of these per keystroke, with the same state. Change-Id: I9d759f54aee8d6dabcef094997e8f352dd608ec3 Reviewed-on: https://gerrit.libreoffice.org/c/core/+/128539 Tested-by: Jenkins Reviewed-by: Michael Meeks <michael.meeks@collabora.com>
2021-12-16ensure invalidate tiles LOK message range checkingLuboš Luňák
3db1ce30ab235ad22aed71c22e4f6f52b7b88829 added some range checking on the rectangles, such as making sure (x,y) is not less than (0,0), and it added it to converting string messages back to rectangles. But then 3b729db05553c1a6d461fb41c89a05702f407448 avoided the conversions from string messages back to rectangles, and thus it avoided also these checks. Change-Id: I73a08e418dc2e48ef5e89fe1aee0272851f7d363 Reviewed-on: https://gerrit.libreoffice.org/c/core/+/126865 Tested-by: Jenkins Reviewed-by: Luboš Luňák <l.lunak@collabora.com>
2021-12-14fix handling of the "EMPTY" LOK tile invalidationLuboš Luňák
The LOK_CALLBACK_INVALIDATE_TILES documentation says that invalidate-all message should say "EMPTY", which wasn't converted properly from the MaxTwips rectangle representation. Doing that now needs also changing the testTileInvalidationCompression() test, but that should be a fix of the test, and conceptually it should be the same. Change-Id: I58fcc56ee56d9f6fcdb9298938e8aa7e3609d6db Reviewed-on: https://gerrit.libreoffice.org/c/core/+/126651 Tested-by: Jenkins Reviewed-by: Luboš Luňák <l.lunak@collabora.com>
2021-12-07make sure vector elements are initialized properlyLuboš Luňák
Change-Id: If18268a26118ea587f474c21b7ca0f1fa4622744 Reviewed-on: https://gerrit.libreoffice.org/c/core/+/126469 Tested-by: Jenkins Reviewed-by: Luboš Luňák <l.lunak@collabora.com>
2021-11-03loplugin:constparamsNoel Grandin
Change-Id: Iebeb531fad5cc819b536788925cf8508737198b7 Reviewed-on: https://gerrit.libreoffice.org/c/core/+/124599 Tested-by: Jenkins Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
2021-10-26use boost flat_map for faster mapLuboš Luňák
setUpdatedTypePerViewId() gets called so often that even unordered_map lookup and allocations show up in profiling. It seems that flat_map can do better. Change-Id: Id2e4f0c40d6973b51e557f84a08bf12feb58b3ce Reviewed-on: https://gerrit.libreoffice.org/c/core/+/124210 Tested-by: Jenkins Reviewed-by: Luboš Luňák <l.lunak@collabora.com>
2021-10-24change some LOK internal updates to be pull model instead of pushLuboš Luňák
Some LOK messages may get called very often, such as updates about cursor position. And since only the last one matters, they get generated every time, which costs some time, and then later except for one they get all discard again from CallbackFlushHandler queue, which again costs time. Change the model to instead only set an 'updated' flag, and CallbackFlushHandler will request the actual message payload only before flushing. This commit changes LOK_CALLBACK_INVALIDATE_VISIBLE_CURSOR and LOK_CALLBACK_INVALIDATE_VIEW_CURSOR to work this way. Change-Id: I376be63176c0b4b5cb492fbf529c21ed01b35481 Reviewed-on: https://gerrit.libreoffice.org/c/core/+/124083 Tested-by: Jenkins Reviewed-by: Luboš Luňák <l.lunak@collabora.com>
2021-10-22better name for a function overloadLuboš Luňák
It's better to add to the name what the overload does rather than just have a "mysterious" extra int. Change-Id: Iff89679c4a978a4596ac662ef74e934cdefefc9e Reviewed-on: https://gerrit.libreoffice.org/c/core/+/124001 Tested-by: Jenkins Reviewed-by: Luboš Luňák <l.lunak@collabora.com>
2021-10-15delay, collect and compress LOK invalidations for Writer viewsLuboš Luňák
Due the to way views are updated on any document change, invalidations are at least O(n^2), and since LOK may use a number of views and for each view the entire document is considered to be the view area, this can lead to a huge number of invalidations that are mostly the same repeated rectangles. Change-Id: I63682958d2fc388344641dcd19fa1d2b39054b51 Reviewed-on: https://gerrit.libreoffice.org/c/core/+/123617 Tested-by: Jenkins Reviewed-by: Luboš Luňák <l.lunak@collabora.com>
2021-10-15do not use text-based LOK callback internallyLuboš Luňák
CallbackFlushHandler post-processes LOK messages, but for things like dropping useless invalidations it needs to know the rectangle or the view id, and since the only data it gets are string messages, it needs to convert those back to binary form. Which is slow with large numbers of messages. Add internal LOK callback variant that allows also passing specific data in the original binary form. And then use directly the binary data in CallbackFlushHandler. Change-Id: I8dd30d2ff9c09feadebc31a44d8e6a8ccc306504 Reviewed-on: https://gerrit.libreoffice.org/c/core/+/123589 Tested-by: Jenkins Reviewed-by: Michael Meeks <michael.meeks@collabora.com>
2021-10-12loplugin:unusedmethodsNoel Grandin
Change-Id: Ifd3a1ccef68ebc4cd4e7785357e6a476f6669eb8 Reviewed-on: https://gerrit.libreoffice.org/c/core/+/123456 Tested-by: Jenkins Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
2021-09-28add extra timeout with higher priority to LOK flushingLuboš Luňák
The normal idle has TaskPriority::POST_PAINT, which means that if we get too busy, the idle won't be called for a long time, meaning the queue will get longer and longer, making its processing slower, and client interactivity will be very poor, with updates possibly coming only when everything becomes idle. The second timeout will flush the queue after a reasonable timeout. I don't think there's an optimal value, so let's choose 100ms for now. Change-Id: Ia1312a690aefd2c8628c82e0f42b2993802d8b1e Reviewed-on: https://gerrit.libreoffice.org/c/core/+/122728 Tested-by: Luboš Luňák <l.lunak@collabora.com> Reviewed-by: Luboš Luňák <l.lunak@collabora.com>
2021-09-28optimize removing from the LOK flush queueLuboš Luňák
All the lambdas check for event type, so it makes sense to first separately check the type and only then possibly call the lambda. Especially since 3b3e4ee97af23f21 separated the types for better searching. Change-Id: I144c88f5319ac2141336e1aa3c4ffd7b38265af9 Reviewed-on: https://gerrit.libreoffice.org/c/core/+/122673 Tested-by: Jenkins Reviewed-by: Luboš Luňák <l.lunak@collabora.com>
2021-09-04speed up scanning the LOK queueNoel Grandin
we frequently scan the queue to caolesce events. Most of the time we are scanning based on the event type. So split the queue data into a compact queue that only contains the type, and another queue for the rest of the data. That makes the scanning __much__ more cache-friendly. Change-Id: I92d0b95611cd139cac8532f9297eaabda71d5fe9 Reviewed-on: https://gerrit.libreoffice.org/c/core/+/119996 Tested-by: Jenkins CollaboraOffice <jenkinscollaboraoffice@gmail.com> Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk> (cherry picked from commit acf9cf33d53e4bf598ddbdab102bfbd6bb14f8a3) Reviewed-on: https://gerrit.libreoffice.org/c/core/+/121558 Tested-by: Jenkins
2021-08-05improve LOK notifyInvalidationNoel Grandin
(*) tweak buffer in SfxLokHelper::notifyInvalidation to be a bit larger, so we avoid the cost of a resize&copy (*) use our optimised OString concatentation instead of going via std::stringstream (*) pass down a pointer to rectangle, instead of a string. later we will use this to avoid doing the stringify until later Change-Id: Ia3e3042bc919d9b9cb80e47a93704eb236438605 Reviewed-on: https://gerrit.libreoffice.org/c/core/+/119994 Tested-by: Jenkins CollaboraOffice <jenkinscollaboraoffice@gmail.com> Reviewed-by: Michael Meeks <michael.meeks@collabora.com> Reviewed-on: https://gerrit.libreoffice.org/c/core/+/120072 Tested-by: Jenkins Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
2020-12-29loplugin:stringviewparam: operator +Stephan Bergmann
Change-Id: I044dd21b63d7eb03224675584fa143009c6b6008 Reviewed-on: https://gerrit.libreoffice.org/c/core/+/108418 Tested-by: Jenkins Reviewed-by: Stephan Bergmann <sbergman@redhat.com>
2020-11-21sfx2: lok: reliably support multi-documentsAshod Nakashian
Instead of using the current view to set the DocId, we instead make sure that the ShellView object has the DocId set at construction time. This turned out to be necessary in at least one case (which has a unit-test that failed), which is when events fired during the creation of a new view. The cursor position is notified before we have a chance to set the DocId and because of that we miss the notifications (or worse, we end up sending them to all other documents' views in an effort to fix this bug). This approach is clean and always guarantees that all views have the correct DocId set as soon as possible and that all notifications are sent as expected. A unit-test is added to exercise mult-document usage, which exposed a number of bugs and issues that have been addressed in this patch. Change-Id: Icf5145fb1dabd0d029368310c2b9bf73ae927ccc Reviewed-on: https://gerrit.libreoffice.org/c/core/+/99975 Tested-by: Jenkins CollaboraOffice <jenkinscollaboraoffice@gmail.com> Reviewed-by: Michael Meeks <michael.meeks@collabora.com> Reviewed-by: Ashod Nakashian <ash@collabora.com> Reviewed-on: https://gerrit.libreoffice.org/c/core/+/106252 Tested-by: Jenkins Reviewed-by: Tor Lillqvist <tml@collabora.com>
2020-09-13tdf#124176 Use #pragma once in desktopGeorge Bateman
This commit was carried out by a Python script, source of which is at https://bugs.documentfoundation.org/show_bug.cgi?id=124176#c97. Change-Id: I26f01467d2a572a51c7ace76628d4a8f96f249a4 Reviewed-on: https://gerrit.libreoffice.org/c/core/+/102553 Tested-by: Jenkins Reviewed-by: Julien Nabet <serval2412@yahoo.fr>
2020-06-09tdf#128502: Try to support multiple documents in LibreOfficeKit-using processTor Lillqvist
The LibreOfficeKit-specific code typically has assumed that all the "views" (SfxViewShell instances) are for the same document, because all LibreOfficeKit-based application processes (including the "kit" processes in Online and the iOS app) so far have only had one document open at a time. It is now possible to pass several document file names on the command line to gtktiledviewer and that is an easy way to demonstrate how badly it still works even with this patch. Introduce a unique numeric document id that is increased in the LibLODocument_Impl constructor. Add APIs to access that. When iterating over views, try to skip views that are not of the document visible in the "current" view, if we know what the "current" view is. Also add a couple of FIXMEs at places where it is a bit unclear (to me) whether it is correct to iterate over all views, or whether only views for the "current" document are what we would want. Change-Id: Id5ebb92a115723cdeb23907163d5b5f282016252 Reviewed-on: https://gerrit.libreoffice.org/c/core/+/95353 Tested-by: Jenkins Reviewed-by: Tor Lillqvist <tml@collabora.com>
2020-01-24tdf#42949 Fix IWYU warnings in desktop/Gabor Kelemen
Found with bin/find-unneeded-includes Only removal proposals are dealt with here. Change-Id: If72377cb352be71050a5b0a471bbe3b02d9c0f74 Reviewed-on: https://gerrit.libreoffice.org/c/core/+/87117 Tested-by: Jenkins Reviewed-by: Miklos Vajna <vmiklos@collabora.com>
2020-01-14loplugin:finalclasses in basic..emfioNoel Grandin
Change-Id: I1b0df1a6cb5b8db9db09cb1d55d932459ab16d81 Reviewed-on: https://gerrit.libreoffice.org/c/core/+/86741 Tested-by: Jenkins Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
2019-12-13loplugin:unusedfieldsNoel Grandin
Change-Id: I28c8d60543d85ea878698a3e3661bbd5891ad9c7 Reviewed-on: https://gerrit.libreoffice.org/85075 Tested-by: Jenkins Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
2019-12-03desktop: disable callback handling while changing the viewAshod Nakashian
When changing the view, some components are disabled and re-enabled. This triggers a flood of invalidations that then result in rendering requests. For each rendering the view is set, which triggers more invalidations. The Sidebar suffers from this, and it causes cpu pegging. This fix prevents this issue by disabling all callbacks during setView. Change-Id: If6b93b2ab31f568a0761f15d945a43de1bc2d4d0 Reviewed-on: https://gerrit.libreoffice.org/84184 Reviewed-by: Ashod Nakashian <ashnakash@gmail.com> Tested-by: Ashod Nakashian <ashnakash@gmail.com> Reviewed-on: https://gerrit.libreoffice.org/84226 Tested-by: Jenkins Reviewed-by: Michael Meeks <michael.meeks@collabora.com>
2019-07-25LOK: Refactor large event queueing casesAshod Nakashian
This makes it more readable and reduces indentations. No functional change intended. (cherry picked from commit 570a3c5361d58635e4d0ac8baf6df787bd2290ad) Change-Id: I129ecff9083c459d633c0965eab6e06ba3dcf499 Reviewed-on: https://gerrit.libreoffice.org/76320 Tested-by: Jenkins Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
2019-07-25LOK: Reduce logging while processing eventsAshod Nakashian
Change-Id: I8ca457387715fcd085bcbf3107839d2629580f7b Reviewed-on: https://gerrit.libreoffice.org/76296 Tested-by: Jenkins Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
2019-02-22LOK: reuse cached json object for ViewId where availableAshod Nakashian
Change-Id: I5e29ec2863e06d3dfcbad95c55e355805f12e259 Reviewed-on: https://gerrit.libreoffice.org/67895 Tested-by: Jenkins Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
2019-02-22LOK: always update the payload string via the cacheAshod Nakashian
Change-Id: I090ceb7d214d0809e86b023a8806b6655627918a Reviewed-on: https://gerrit.libreoffice.org/67894 Tested-by: Jenkins Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
2019-02-22LOK: Validate cached queued callback payloadsAshod Nakashian
And with DBG_UTIL validate and dump LOK queue state. Change-Id: I211ddf89dbcecf17c4f1401e5d96919c4bc966ea Reviewed-on: https://gerrit.libreoffice.org/67893 Tested-by: Jenkins Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
2019-02-22LOK: Cache JSON payloadsAshod Nakashian
Change-Id: I81f74027363d4b2959c053057851cf01fc94af8b Reviewed-on: https://gerrit.libreoffice.org/67892 Tested-by: Jenkins Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
2019-02-22LOK: Cache RectangleAndPart payloadsAshod Nakashian
Change-Id: I3a2d98426729ad1b4e43a8f657b512679be82c26 Reviewed-on: https://gerrit.libreoffice.org/67891 Tested-by: Jenkins Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
2019-02-22LOK: Use a struct instead of std::pair to support cachingAshod Nakashian
Change-Id: I48e8de8709e017f2667c0bf2c850004128c840d1 Reviewed-on: https://gerrit.libreoffice.org/67890 Tested-by: Jenkins Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
2019-02-22LOK: Move RectangleAndPart to the headerAshod Nakashian
This is in preparation to cache them. Change-Id: Ic511caf5a8798750288e9271f6898ab38fe2055f Reviewed-on: https://gerrit.libreoffice.org/67889 Tested-by: Jenkins Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
2018-06-20lok: Unit test for jsonToPropertyValuesVector.Jan Holesovsky
Change-Id: I3e0623cc68838c650edbd03cc89bf3fcb8098ff8 Reviewed-on: https://gerrit.libreoffice.org/56149 Tested-by: Jenkins Reviewed-by: Jan Holesovsky <kendy@collabora.com>
2017-06-12cleanup unused css/frame/* includesJochen Nitschke
Change-Id: I173a29fd1ee889127369d2bc2fce8e010b89ca65 Reviewed-on: https://gerrit.libreoffice.org/38633 Tested-by: Jenkins <ci@libreoffice.org> Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
2017-03-29Fix typosAndrea Gelmini
Change-Id: I379c3f8ca971e6180744fb7270dde9164ebcec1a Reviewed-on: https://gerrit.libreoffice.org/35814 Tested-by: Jenkins <ci@libreoffice.org> Reviewed-by: Julien Nabet <serval2412@yahoo.fr>
2017-03-28Fix typosAndrea Gelmini
Change-Id: Ibef75de246fb31884153437fe8c05e299fe5962f Reviewed-on: https://gerrit.libreoffice.org/35782 Tested-by: Jenkins <ci@libreoffice.org> Reviewed-by: Michael Stahl <mstahl@redhat.com>
2017-03-24lok: Allow setting of the language during load.Jan Holesovsky
Change-Id: I9dbb62950e639376c26122ceb9fcec2982b3ca82
2017-01-31Revert "Lok: support for batch API calls"Jan Holesovsky
This was for the moment removed from the online.git, so let's disable the API bits too before it shows this is really necessary. Reverts also: Revert "Lok: unittest batch API" Change-Id: I7bf3fe62d1e73b6f233992d51f587868a78f4bec
2017-01-23Lok: support for batch API callsAshod Nakashian
Mouse and keyboard operations typically come in batches, and often each results in tile invalidations and/or layout modifications. Processing each input event on its own, then processing the resulting output event is very costly and unecessary when we know there is more of the same. The new API adds support for batching such related input events by disabling the output events generated by Core until the batch is done. The client can then process the resulting events, which will be compressed and deduplicated. Change-Id: Id381dab807186d010021a8778ee440074a739920 Reviewed-on: https://gerrit.libreoffice.org/33402 Tested-by: Jenkins <ci@libreoffice.org> Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
2016-11-09loplugin:expandablemethods in dbaccessNoel Grandin
Change-Id: I6087a3eff46926646ac1637615a0af30b38956a4 Reviewed-on: https://gerrit.libreoffice.org/30712 Tested-by: Jenkins <ci@libreoffice.org> Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
2016-09-13loplugin:override: No more need for the "MSVC dtor override" workaroundStephan Bergmann
The issue of 362d4f0cd4e50111edfae9d30c90602c37ed65a2 "Explicitly mark overriding destructors as 'virtual'" appears to no longer be a problem with MSVC 2013. (The little change in the rewriting code of compilerplugins/clang/override.cxx was necessary to prevent an endless loop when adding "override" to OOO_DLLPUBLIC_CHARTTOOLS virtual ~CloseableLifeTimeManager(); in chart2/source/inc/LifeTime.hxx, getting stuck in the leading OOO_DLLPUBLIC_CHARTTOOLS macro. Can't remember what that isAtEndOfImmediateMacroExpansion thing was originally necessary for, anyway.) Change-Id: I534c634504d7216b9bb632c2775c04eaf27e927e
2016-09-10LOK: new callback dropping implementationMarco Cecchetti
Now view callbacks have their own collection of last states where the key is made up by both the view id and the callback type. Callback dropping based on the last state is no more handled on queueing but on flushing, since what really matters is the last performed callback (for each callback type). Anyway in order to not modify the order of callbacks, that could be changed when an already queued callback is superseeded, dropping still occurs on queuing too, just by looking for the last queued callback of the same type. The result is a substantial reduction of redundant callbacks and fix the following problem in loleaflet: when there are more views for a speadsheet and cell cursors for two view are placed on the same cell, a continuos swapping between the two cell cursors can occur. That was due to a sequence of "EMPTY" and coordinates messages or cell cursor and cell view cursor messages which were sent in an alternating way. Change-Id: I79e14d11d4e8590aff715181e3410ad88c4e6175 Reviewed-on: https://gerrit.libreoffice.org/28783 Tested-by: Jenkins <ci@libreoffice.org> Reviewed-by: Marco Cecchetti <mrcekets@gmail.com>
generated by cgit (git 2.34.1) at 2025-01-31 17:08:53 +0000