Development process problem: the self-dependency loop in the compiler is too tight

[johnwcowan]

Today I tried to recompile the current head after a make confclean, and found that my copy of chicken dating from 2011-09-21 (11 days old) was unable to compile it, because it did not understand the specialization type input-port. It's reasonable to add a new specialization type. It's not reasonable to change the source to depend on it at the same time. Now I'm stuck: until there is a new release tarball from which I can get an up-to-date chicken, I can't rebuild my system.

I see several possibilities here.

1. Use a new branch for fast-changing but un-bypassable features like specialization. I am basically in favor of doing all development on trunk, but because Chicken builds depend so delicately on Chicken builds, some concessions may have to be made.

2. Have a switch to disable optimizations like this -- I tried setting optimize-level to 0 in defaults.make, but it didn't help.

3. As a matter of social engineering, make sure to bump the minimal version number every time a backward incompatibility like this happens, and issue a release tarball. At present the latest tarball is 4.7.2, so old as to be useless with trunk.

[Jim Ursetto]

Did you make boot-chicken first?

[johnwcowan]

I've now read the comments on #696, so closing that one and keeping this as a doc issue.

[johnwcowan]

With the new chickenista page, I'll close this too.

[felix winkelmann]

Milestone 4.8.0 deleted