Using generate_export_header instead manually written export headers
should improve coverage across platforms.
Size of stripped libmarblewidget-qt5 reduced on Linux
e.g. from 11077592 to 7889848
nienhueser | |
rahn |
Marble |
Using generate_export_header instead manually written export headers
should improve coverage across platforms.
Size of stripped libmarblewidget-qt5 reduced on Linux
e.g. from 11077592 to 7889848
No Linters Available |
No Unit Test Coverage |
Yes, at least I very much assume that the compiler now can optimize more code due to no longer having to support the public state of all symbols. No experience in putting binaries under the microscope yet for that, sadly.
All I could see/feel, next to size of lib binary (those numbers were for stripped version) was a slightly improved start-up time, which surely is due to dynamic linker having less symbols to walk through :)
Meh. For some reason this results in a broken apk, at least on my SFOS with its Dalvik adaption.
Will at least commit removal of long-time deprecated methods, and look more later into how to get access to any logs of android apps on my sfos. If anyone with proper Android emulator/device could help and have a look as well, would be welcome.
Ah, that is the reason, the marblemaps "executable" for Android is built as shared lib, so that needs proper handling:
D/dalvikvm( 2793): Trying to load lib /data/data/org.kde.marble.maps/lib/libMarble.so 0x418d5558 D/dalvikvm( 2793): Added shared lib /data/data/org.kde.marble.maps/lib/libMarble.so 0x418d5558 D/dalvikvm( 2793): No JNI_OnLoad found in /data/data/org.kde.marble.maps/lib/libMarble.so 0x418d5558, skipping init E/ ( 2793): (null):0 ((null)): dlsym failed: Symbol not found: E/ ( 2793): (null):0 ((null)): Could not find main method