GSoC/2021/StatusReports/AnjaniKumar: Difference between revisions
No edit summary |
No edit summary |
||
| Line 70: | Line 70: | ||
== Coding Period (June 7 - August 16)== | == Coding Period (June 7 - August 16)== | ||
=== Week 1 === | |||
I patched a lot of warnings mostly from Qt classes that had fewer warnings. I also wrote a new unit test for FilterAction class. This test was related to ''QHash'' warnings. The test checked if all keys are unique in a QHash in FilterAction class. | I patched a lot of warnings mostly from Qt classes that had fewer warnings. I also wrote a new unit test for FilterAction class. This test was related to ''QHash'' warnings. The test checked if all keys are unique in a QHash in FilterAction class. | ||
In a nutshell, there were 6693 lines of build warnings on master.With the patches on my branch these warnings reduced to 1830 lines. That’s a reduction of about 73%. | In a nutshell, there were 6693 lines of build warnings on master.With the patches on my branch these warnings reduced to 1830 lines. That’s a reduction of about 73%. | ||
==== Related Patches ==== | |||
* [https://invent.kde.org/graphics/digikam/-/merge_requests/84 Fix QMap::unite() deprecation warning in presentation plugin ] | * [https://invent.kde.org/graphics/digikam/-/merge_requests/84 Fix QMap::unite() deprecation warning in presentation plugin ] | ||
| Line 85: | Line 83: | ||
* [https://invent.kde.org/graphics/digikam/-/merge_requests/91 New Unit test: FilterAction Unique keys test for all filters] | * [https://invent.kde.org/graphics/digikam/-/merge_requests/91 New Unit test: FilterAction Unique keys test for all filters] | ||
==== Related Blog ==== | |||
https://anjani.live/posts/gsoc21-week-1/ | https://anjani.live/posts/gsoc21-week-1/ | ||
=== Week 2 === | |||
Continuation of work from previous week. We also discussed about introducing STL algorithms in digiKam. Though it is huge task in its own and not related to my project so we suppressed the ''cppcheck'' warnings in report and moved on. | Continuation of work from previous week. We also discussed about introducing STL algorithms in digiKam. Though it is huge task in its own and not related to my project so we suppressed the ''cppcheck'' warnings in report and moved on. | ||
==== Related Patches ==== | |||
* [https://invent.kde.org/graphics/digikam/-/merge_requests/95 Fix QSet issues by using only QSet constructor to make QSet objects] | * [https://invent.kde.org/graphics/digikam/-/merge_requests/95 Fix QSet issues by using only QSet constructor to make QSet objects] | ||
| Line 103: | Line 101: | ||
* [https://invent.kde.org/graphics/digikam/-/commit/280a47d9e7ced1d1a84adba8987cddaab9768e1f Macro for Qt::SplitBehavior] | * [https://invent.kde.org/graphics/digikam/-/commit/280a47d9e7ced1d1a84adba8987cddaab9768e1f Macro for Qt::SplitBehavior] | ||
==== Related Blog ==== | |||
https://anjani.live/posts/gsoc21-week-2/ | https://anjani.live/posts/gsoc21-week-2/ | ||
=== Week 3 === | |||
During this week I patched the very old random number generators and ''qsrand()/qrand()'' in digiKam. As a result we were able to drop ''Boost::Random dependency''. I also ported ''Qt X11 Extras'' module code. A very detailed report is in the blog post. | During this week I patched the very old random number generators and ''qsrand()/qrand()'' in digiKam. As a result we were able to drop ''Boost::Random dependency''. I also ported ''Qt X11 Extras'' module code. A very detailed report is in the blog post. | ||
==== Related patches ==== | |||
* [https://invent.kde.org/graphics/digikam/-/merge_requests/104 Remove qsrand] | * [https://invent.kde.org/graphics/digikam/-/merge_requests/104 Remove qsrand] | ||
| Line 119: | Line 117: | ||
* [https://invent.kde.org/graphics/digikam/-/merge_requests/108 Remove deprecated QMutex usage] | * [https://invent.kde.org/graphics/digikam/-/merge_requests/108 Remove deprecated QMutex usage] | ||
==== Related Blog ==== | |||
https://anjani.live/posts/gsoc21-week-3/ | https://anjani.live/posts/gsoc21-week-3/ | ||
=== Week 4 === | |||
In this week, digiKam was fully ported to Qt 5.15. We also started discussion on now introducing Qt6 this week. I tried building it for the first time with Qt6, it failed(no surprises here!). We ran into a problem that digiKam depends on Qt WebEngine which would only be stable in Qt 6.2 in September. We discussed about possible solutions. I tried building the alpha Qt 6.2 but it doesn't build. | In this week, digiKam was fully ported to Qt 5.15. We also started discussion on now introducing Qt6 this week. I tried building it for the first time with Qt6, it failed(no surprises here!). We ran into a problem that digiKam depends on Qt WebEngine which would only be stable in Qt 6.2 in September. We discussed about possible solutions. I tried building the alpha Qt 6.2 but it doesn't build. | ||
==== Related Patches ==== | |||
* [https://invent.kde.org/graphics/digikam/-/merge_requests/110 Replace removed QDateTime constructor with new function] | * [https://invent.kde.org/graphics/digikam/-/merge_requests/110 Replace removed QDateTime constructor with new function] | ||
| Line 133: | Line 131: | ||
* [https://invent.kde.org/graphics/digikam/-/merge_requests/112 Isolate rajce plugin] | * [https://invent.kde.org/graphics/digikam/-/merge_requests/112 Isolate rajce plugin] | ||
=== Week 5 === | |||
I found the solution to problem from last week. Gilles also helped me to get started. This week I worked on porting the build system to Qt6. digiKam is huge and we have successfully branched the build process if no web module is found. A more detailed report is in the blog. | I found the solution to problem from last week. Gilles also helped me to get started. This week I worked on porting the build system to Qt6. digiKam is huge and we have successfully branched the build process if no web module is found. A more detailed report is in the blog. | ||
==== Related Patches ==== | |||
* [https://invent.kde.org/graphics/digikam/-/merge_requests/114 Dummy webwidget] | * [https://invent.kde.org/graphics/digikam/-/merge_requests/114 Dummy webwidget] | ||
==== Related Blog ==== | |||
https://anjani.live/posts/gsoc21-week-5/ | https://anjani.live/posts/gsoc21-week-5/ | ||
=== Week 6 === | |||
This week our goal was to port the build system in such a way that it can detect a Qt installation 5/6 and adjust accordingly. Our goal was to preserve the functionality of current Qt5 pipelines and tests. We used cmake variables and patched hardcoded Qt versions in the cmake with dynamically set variables enabling cmake configuration with both Qt5 and Qt6. | This week our goal was to port the build system in such a way that it can detect a Qt installation 5/6 and adjust accordingly. Our goal was to preserve the functionality of current Qt5 pipelines and tests. We used cmake variables and patched hardcoded Qt versions in the cmake with dynamically set variables enabling cmake configuration with both Qt5 and Qt6. | ||
==== Related Patches ==== | |||
* [https://invent.kde.org/graphics/digikam/-/merge_requests/119 Port CMake files] | * [https://invent.kde.org/graphics/digikam/-/merge_requests/119 Port CMake files] | ||
| Line 157: | Line 155: | ||
and many more can be found in the work branch. | and many more can be found in the work branch. | ||
==== Related Blog ==== | |||
https://anjani.live/posts/gsoc21-week-6/ | https://anjani.live/posts/gsoc21-week-6/ | ||
=== Week 7 === | |||
This week I worked on building digiKam with Qt6. To do this, I had to build KF5 components with Qt6 first as they are currently built with Qt5 which was a problem. I built a few modules with Qt6 but it was taking some time. If I had more time in my hands I could have completed building all required modules but we decided that we can do this later and move to work on the next task. | This week I worked on building digiKam with Qt6. To do this, I had to build KF5 components with Qt6 first as they are currently built with Qt5 which was a problem. I built a few modules with Qt6 but it was taking some time. If I had more time in my hands I could have completed building all required modules but we decided that we can do this later and move to work on the next task. | ||
| Line 167: | Line 165: | ||
Now I had to port ''Core5Compat'' module code to ''Qt6 Core'' code. I have explained all the work in detail in the blog post. | Now I had to port ''Core5Compat'' module code to ''Qt6 Core'' code. I have explained all the work in detail in the blog post. | ||
==== Related Patches ==== | |||
* [https://invent.kde.org/graphics/digikam/-/merge_requests/124 Port QRegExp to QRegularExpression in dplugins] | * [https://invent.kde.org/graphics/digikam/-/merge_requests/124 Port QRegExp to QRegularExpression in dplugins] | ||
| Line 175: | Line 173: | ||
* [https://invent.kde.org/graphics/digikam/-/merge_requests/129 Port QRegExp::exactMatch()] | * [https://invent.kde.org/graphics/digikam/-/merge_requests/129 Port QRegExp::exactMatch()] | ||
==== Related Blog ==== | |||
https://anjani.live/posts/gsoc21-week-7/ | https://anjani.live/posts/gsoc21-week-7/ | ||
=== Week 8 === | |||
Most of this week's work was debugging the regression in Porting QRegExp to QRegularExpression in ''advancedrename'' tool. The regression was caught by ''advancedrename_utest''. The regression was fixed by patching the advancedrename digiKam API itself. I have explained the regression and my process of fixing in the blog post so do refer that. | Most of this week's work was debugging the regression in Porting QRegExp to QRegularExpression in ''advancedrename'' tool. The regression was caught by ''advancedrename_utest''. The regression was fixed by patching the advancedrename digiKam API itself. I have explained the regression and my process of fixing in the blog post so do refer that. | ||
| Line 190: | Line 188: | ||
For the time being, we will use ''Core5Compat'' module to keep this code working. | For the time being, we will use ''Core5Compat'' module to keep this code working. | ||
==== Related Patches ==== | |||
* [https://invent.kde.org/graphics/digikam/-/merge_requests/130 Port QRegExp in advancedrename tool] | * [https://invent.kde.org/graphics/digikam/-/merge_requests/130 Port QRegExp in advancedrename tool] | ||
* [https://invent.kde.org/graphics/digikam/-/merge_requests/131 Port QTextCodec] | * [https://invent.kde.org/graphics/digikam/-/merge_requests/131 Port QTextCodec] | ||
==== Related Blog ==== | |||
https://anjani.live/posts/gsoc21-week8/ | https://anjani.live/posts/gsoc21-week8/ | ||
=== Week 9 === | |||
This week, I ported the remaining '''QRegExp''' uses in digiKam. There was another regression detected this week as well in class ''cameranamehelper''. The cause was not because of my mistake. We think that it is probably a bug in Qt. This regex pattern "''^(ptp|normal|mtp)(\\s+mode)?$''" was failing to give an exact match with '''QRegularExperssion''' class in ''case insensitive'' mode. I found a way to make it work by using a different construction style. The blog post has detailed information. | This week, I ported the remaining '''QRegExp''' uses in digiKam. There was another regression detected this week as well in class ''cameranamehelper''. The cause was not because of my mistake. We think that it is probably a bug in Qt. This regex pattern "''^(ptp|normal|mtp)(\\s+mode)?$''" was failing to give an exact match with '''QRegularExperssion''' class in ''case insensitive'' mode. I found a way to make it work by using a different construction style. The blog post has detailed information. | ||
==== Related Patches ==== | |||
* [https://invent.kde.org/graphics/digikam/-/merge_requests/132 Port remaining QRegExp uses] | * [https://invent.kde.org/graphics/digikam/-/merge_requests/132 Port remaining QRegExp uses] | ||
==== Related Blog ==== | |||
https://anjani.live/posts/gsoc21-week9-10/ | https://anjani.live/posts/gsoc21-week9-10/ | ||
=== Week 10 === | |||
In the final week of the coding period, I worked on the appimage builder scripts. There are '''4''' scripts and '''4'''' helper scripts to build appimages. The 4 main scripts are: | In the final week of the coding period, I worked on the appimage builder scripts. There are '''4''' scripts and '''4'''' helper scripts to build appimages. The 4 main scripts are: | ||
| Line 222: | Line 220: | ||
Because earlier we had decided to leave building KF5 with Qt6 for later, I could only work on the 1st script for now. My task was to patch the script so that it can build Qt '''6.2.0-beta2''' on '''Mageia 7.1'''. The script is yet to be tested fully. | Because earlier we had decided to leave building KF5 with Qt6 for later, I could only work on the 1st script for now. My task was to patch the script so that it can build Qt '''6.2.0-beta2''' on '''Mageia 7.1'''. The script is yet to be tested fully. | ||
==== Related Patches ==== | |||
* [https://invent.kde.org/graphics/digikam/-/merge_requests/138 Draft: Build host with Qt6 for appimage] | * [https://invent.kde.org/graphics/digikam/-/merge_requests/138 Draft: Build host with Qt6 for appimage] | ||
==== Related Blog ==== | |||
https://anjani.live/posts/gsoc21-week9-10/ | https://anjani.live/posts/gsoc21-week9-10/ | ||
Revision as of 12:28, 20 August 2021
digiKam: Port to Qt6 on Linux
digiKam is a professional photo management software. With the release of Qt6, digiKam needs to to adapt to many new improvements made possible with the new Qt6 framework such as code readability, better classes from Qt6 and new unit tests.
Mentors
- Gilles Caulier
- Maik Qualmann
- Thanh Trung Dinh
Links
Goals
- To prepare digiKam for Qt6 stable release in future.
- Test for regressions and fix them
- Write new AppImage builder scripts for future Qt6 release
Community Bonding (May 18 - June 6)
We discussed the proposal again and made some strategies on the first task and how to test the new patches.The first task was to Port digiKam to Qt 5.15. This is a sophisticated task and involves a lot of sub-steps like:
- Identifying deprecated classes and class members/functions in Qt 5.15
- Identifying their uses in digiKam code and organizing them based on which class has most uses.
- Porting the identified code. Sometimes trivial, sometimes needs reviews from mentors.
- Writing new unit test if needed.
- Running tests and check the patched code at runtime with the digiKam stable release.
We planned a safe way to introduce patches safely in the code, otherwise regressions might be introduced. Here is a list that I made of deprecated code in digiKam and sorted it according to number of usages it had.
- QTextStream::endl
- QStringList
- qrand()
- qsrand()
- QSet
- QImage
- QWheelEvent
- QMap
- QHash
- QDateTime
- QButtonGroup
- QPixmap
- QPrinter
- QProcess
- QTabletEvent
- QTimeLine
- QMutex
After identifying and listing the code that needed to be patched, I proposed two methods to approach the problem:
- We pick a source file in digiKam and make patches to it resolving all Qt deprecation warnings.
- We pick a Qt class from our list and identify it's use in all digiKam code and patch them.
The latter was safer as dealing with a lot of Qt classes in the same file might have led to regressions which were difficult to track down.
So I started early in this time only so that I don't get stuck in the coding period. I started picking Qt classes with few usages and started porting them. These were QMap, QPrinter. I faced some difficulty but my mentor's feedback helped me a lot and was able to learn quickly. More about this period in the blog post.
Related blog
https://anjani.live/posts/gsoc-community-bonding/
Coding Period (June 7 - August 16)
Week 1
I patched a lot of warnings mostly from Qt classes that had fewer warnings. I also wrote a new unit test for FilterAction class. This test was related to QHash warnings. The test checked if all keys are unique in a QHash in FilterAction class. In a nutshell, there were 6693 lines of build warnings on master.With the patches on my branch these warnings reduced to 1830 lines. That’s a reduction of about 73%.
Related Patches
- Fix QMap::unite() deprecation warning in presentation plugin
- Fix QProcess deprecated warning in cameracontroller backend
- Fix QHash deprecated declarations in FilterAction class
- Fix QImage deprecated transformed() usages
- New Unit test: FilterAction Unique keys test for all filters
Related Blog
https://anjani.live/posts/gsoc21-week-1/
Week 2
Continuation of work from previous week. We also discussed about introducing STL algorithms in digiKam. Though it is huge task in its own and not related to my project so we suppressed the cppcheck warnings in report and moved on.
Related Patches
- Fix QSet issues by using only QSet constructor to make QSet objects
- Version check for QButtonGroup
- Version check for QWheelEvent
- Version check for QLabel
- Version check for QFlag
- Macro for Qt::endl
- Macro for Qt::SplitBehavior
Related Blog
https://anjani.live/posts/gsoc21-week-2/
Week 3
During this week I patched the very old random number generators and qsrand()/qrand() in digiKam. As a result we were able to drop Boost::Random dependency. I also ported Qt X11 Extras module code. A very detailed report is in the blog post.
Related patches
- Remove qsrand
- Remove qrand
- Port class RandomNumberGenerator to drop boost::random
- Remove Qt X11Extras module
- Remove deprecated QMutex usage
Related Blog
https://anjani.live/posts/gsoc21-week-3/
Week 4
In this week, digiKam was fully ported to Qt 5.15. We also started discussion on now introducing Qt6 this week. I tried building it for the first time with Qt6, it failed(no surprises here!). We ran into a problem that digiKam depends on Qt WebEngine which would only be stable in Qt 6.2 in September. We discussed about possible solutions. I tried building the alpha Qt 6.2 but it doesn't build.
Related Patches
- Replace removed QDateTime constructor with new function
- Port dbjobsthread.cpp to not use deprecated const_iterator “+” operator
- Isolate rajce plugin
Week 5
I found the solution to problem from last week. Gilles also helped me to get started. This week I worked on porting the build system to Qt6. digiKam is huge and we have successfully branched the build process if no web module is found. A more detailed report is in the blog.
Related Patches
Related Blog
https://anjani.live/posts/gsoc21-week-5/
Week 6
This week our goal was to port the build system in such a way that it can detect a Qt installation 5/6 and adjust accordingly. Our goal was to preserve the functionality of current Qt5 pipelines and tests. We used cmake variables and patched hardcoded Qt versions in the cmake with dynamically set variables enabling cmake configuration with both Qt5 and Qt6.
Related Patches
and many more can be found in the work branch.
Related Blog
https://anjani.live/posts/gsoc21-week-6/
Week 7
This week I worked on building digiKam with Qt6. To do this, I had to build KF5 components with Qt6 first as they are currently built with Qt5 which was a problem. I built a few modules with Qt6 but it was taking some time. If I had more time in my hands I could have completed building all required modules but we decided that we can do this later and move to work on the next task.
Now I had to port Core5Compat module code to Qt6 Core code. I have explained all the work in detail in the blog post.
Related Patches
- Port QRegExp to QRegularExpression in dplugins
- Port QRegExp to QRegularExpression in libs/database
- Port QRegExp to QRegularExpression in core/libs
- Port QStringRef to QStringView
- Port QRegExp::exactMatch()
Related Blog
https://anjani.live/posts/gsoc21-week-7/
Week 8
Most of this week's work was debugging the regression in Porting QRegExp to QRegularExpression in advancedrename tool. The regression was caught by advancedrename_utest. The regression was fixed by patching the advancedrename digiKam API itself. I have explained the regression and my process of fixing in the blog post so do refer that.
Another patch was the porting of QTextCodec class. It was trivial. I had to leave two instances because of no clear alternatives were available. These are:
For the time being, we will use Core5Compat module to keep this code working.
Related Patches
Related Blog
https://anjani.live/posts/gsoc21-week8/
Week 9
This week, I ported the remaining QRegExp uses in digiKam. There was another regression detected this week as well in class cameranamehelper. The cause was not because of my mistake. We think that it is probably a bug in Qt. This regex pattern "^(ptp|normal|mtp)(\\s+mode)?$" was failing to give an exact match with QRegularExperssion class in case insensitive mode. I found a way to make it work by using a different construction style. The blog post has detailed information.
Related Patches
Related Blog
https://anjani.live/posts/gsoc21-week9-10/
Week 10
In the final week of the coding period, I worked on the appimage builder scripts. There are 4 scripts and 4' helper scripts to build appimages. The 4 main scripts are:
- Build host (Builds Qt)
- Build extra libs (Builds KF5)
- Build digikam
- Build appimage
Because earlier we had decided to leave building KF5 with Qt6 for later, I could only work on the 1st script for now. My task was to patch the script so that it can build Qt 6.2.0-beta2 on Mageia 7.1. The script is yet to be tested fully.
Related Patches
Related Blog
https://anjani.live/posts/gsoc21-week9-10/
Final Status
Work in progress
Contact
Feel free to get in touch
Anjani Kumar
Email : [email protected]
Telegram: https://t.me/anjanik012
Matrix : @anjanik012:kde.org