It’s been almost one year now that I can’t compile most Arduino sketches because GCC crashes with an internal bug (see bug 105753 – [avr] ICE: in add_clobbers, at config/avr/avr-dimode.md:2705). It is known although the current status is “UNCONFIRMED”, as you can see. I am an active member on Arduino Discord server and I am often unable to provide help because of that. I am not the only one in that situation and GCC maintainers don’t seem to see any urgency…
The latest version of GCC known to work is 10.1.0. I’d like to gently ask if it’s possible to make arduino-ide package depend on version 10.1.0 of avr-gcc instead of the buggy version 12 (as per Manjaro’s current dependency tree). This current situation is a total nightmare, really. Can Manjaro maintainers di something about that?
Thanks in advance.
P.S.: As if I didn’t have any issues right now, I couldn’t create a new post from that very message. The only way I could create a post was to type some junk and then edit my post. I just hope it’s the only time I’ll have run into that issue here…
Arduino package is inherited from Arch Linux. Same with the adruino-avr-core that depends on avr-gcc 12.2.0-1 that depends on the gcc-libs 12.2.1-1 (on Manjaro unstable branch and Arch linux stable branch).
What message? If in doubt about the forum, please see:
I was plainly explaining the source of those packages. I would not advise to go on Arch Linux and report issues if you are not on a Arch Linux installation.
One thing i would make sure tho, is that sketches are build as described here Arduino - ArchWiki
I’m sure once the bug report you shared will have more reports, it will be addressed.
You don’t understand: the bug is not limited to Arduino sketches. All AVR code is potentially hit by it and I for sure was with AVR projects I designed. Besides there’s nothing to do to “make sure sketches are compiled properly”…it is completely irrelevant. It’s an internal GCC bug that’s caused by a regression, as the bug report clearly states.
Consequently, Manjaro’s application tree includes a package that is broken. And, as I tested, avr-gcc-10.1.0 doesn’t exhibit that buggy behaviour. It just confirms what the bug report says. The question is: are Manjaro/Arch maintainers wilfully going to leave a package that is known to be broken for months and critical for embedded development?