I remember Atmel mentioned opensource and AVRDUDE in an official ATMEL power point presentation on AVRs.
It stuck in my head because it was pretty funny to see the word "DUDE" in an official presentation
And still the avrdude maintainer is begging Atmel for ages to get complete, correct protocol documentation. Which Atmel doesn't supply. E.g. all the Xmega support in avrdude happenes only because someone reverse enginnered parts of the protocol and provided the information to the avrdude maintainer.
Debugging support (via avrice) is even worse. It took people thee years to partly support Xmega debugging, and again only after someone worked very hard to reverse engineer and analyze the protocol. And only for some version of Atmel tools. Atmel did shit to help.
Atmel is very quick to claim there is free software support when it helps to sell their microcontrollers. But they almost do nothing to build and maintain it.
I had the impression that Atmel somehow helped the creation of avr-gcc.
That was entirely a community effort. Occasionally with Atmel employees helping out in their spare time, without the support from Atmel.
When Atmel took avr-gcc to make it into part of Toolchain they missed some important patches in their build, but also added own code. I.e. adding code generation support for some AVR versions not supported by avr-gcc at that time.
And that was when it got very ugly. If you enhance GPL code you are required to publish your changes under the GPL, too. Atmel didn't, citing internal organizational issues and all sorts of other excuses. They were outright doing a copyright violation, because it was more convenient for them. I haven't check if they finally came to their senses. Maybe they have finally published their changes.
But the fact remains that Atmel is withholding important information needed by open source tools, and that Atmel doesn't play by the rules. What they like to do, however, is to pretend they do all the AVR free software work and get credit for it when it helps sales.