Archimage 4.3

[ivan-hc]

• select only needed dependencies and libraries
• restarting process does not bloats the AppDir
• JuNest will be keept as is in a dedicated directory, while the AppDir will be another directory
• workflow lasts less
• less resource intensive
• reduced functions to check libraries
• AppImages will be cleaner, without empty directories under /usr (eccept mountpoints)
• message to advertice the user that the script is saved in the current directory, in case there is no XDG Desktop directory
• add new options --devmode-enable and --devmode-disable to use the "dev" branch
• add message about "Developer Mode" if enabled
• if "AM" is installed, use bash-completion / zsh-completion

[ivan-hc]

This release has a regression: levels 2, 3 or 4 of extraction just made the apps bloated, but the core libraries to launch the app (that were enough with a level 2 in the version 4.2) have not been included. In that case, the app can also run with a default level 1, but you have to add manually in LIBSAVED the needed keywords for dependencies.

I have merged a lot of functions for libraries into one

_savelibs() {
	echo "◆ Detect libraries related to /usr/bin files"
	readelf -d ./"$APP".AppDir/.junest/usr/bin/* 2>/dev/null | grep NEEDED | tr '[] ' '\n' | grep ".so" | uniq >> ./list

	echo "◆ Detect libraries of the main package"
	base_libs=$(find ./base -type f | uniq)
	for b in $base_libs; do
		readelf -d "$b" 2>/dev/null | grep NEEDED | tr '[] ' '\n' | grep ".so" | uniq >> ./list &
	done
	wait

	echo "◆ Detect libraries of the dependencies"
	dep_libs=$(find ./deps -executable -type f | uniq)
	for d in $dep_libs; do
		readelf -d "$d" 2>/dev/null | grep NEEDED | tr '[] ' '\n' | grep ".so" | uniq >> ./list &
	done
	wait

	echo "◆ Saving libraries"
	cp -r ./archlinux/.junest/usr/lib/ld-linux-x86-64.so* ./"$APP".AppDir/.junest/usr/lib/
	lib_preset="$APP $BIN libdw libelf $(sort -u ./list)"
	LIBSAVED="$lib_preset $LIBSAVED $SHARESAVED"
	for arg in $LIBSAVED; do
		LIBPATHS="$LIBPATHS $(find ./archlinux/.junest/usr/lib -maxdepth 20 -wholename "*$arg*" | sed 's/\.\/archlinux\///g')"
	done
	for arg in $LIBPATHS; do
		cp -r ./archlinux/"$arg" "$APP".AppDir/"$arg" 2>/dev/null &
	done
	wait
	rm list
}

libraries are added directly to the AppDir.

I'll be honest, I don't want to spit the libraries part to several functions again... but also I don't want this regression.

To do an example, for Archimage 4.2 and without including all dependencies, this was LIBSAVED in Abiword

LIBSAVED="gconv db Graphics libical libxslt libXt $lib_audio_keywords $lib_browser_launcher"

and this is it now

LIBSAVED="gconv db Graphics libical libxslt libXt libsqlite libijs- libjbig libidn libpaper.so libopenjp libnghttp libssh \
libpsl.so libssl.so libcrypto.so libgssapi_krb5 libbrotlicommon.so libSM.so libICE.so libunistring.so libkrb libk5crypto libcom_err.so \
libkeyutils.so libtasn libhogweed.so libnettle.so $lib_audio_keywords $lib_browser_launcher"

...and yes, the AppImage is 2 MB smaller... but I worked a lot of time to reach this result, by analyzing the breaks at each launch, library by library, using LD_DEBUG

I need to fix this regression asap.

[ivan-hc]

I have solved!

Now the whole process is faster and less resource-intensive!

I'm experimenting this on some of my repos, and workflows last 1 minute less if compared with Archimage 4.2!