Steve Langasek <[EMAIL PROTECTED]> writes: > On Mon, Jun 05, 2006 at 04:47:44PM -0700, Thomas Bushnell BSG wrote: >> The correct response, I believe, is for Debian guile-1.6 to declare in >> the architecture spec that ia64 is not supported. > >> This is also a release engineering issue, because it means that >> packages which use guile (say, like gnucash) would need to be able to >> slide into testing despite not being functional on ia64. > >> It seems like it would be incorrect to declare in gnucash that ia64 is >> not supported, because there is no gnucash-specific problem. Still, I >> don't understand all the release engineering issues. > >> Steve, can you offer some suggestions? Assume, for the moment, that >> it will not be possible to make guile work on ia64. What then? > > If guile-1.6 declares, with the consent of the ia64 porters, that ia64 is > not supported (either by explicitly excluding it from the architecture list > or by adding a test suite that fails at build-time), then it would be > appropriate to ask the ftp team to drop the ia64 binaries of gnucash from > unstable. The procedure for doing this is to file a bug report against > ftp.debian.org.
Great. So the problem, by the way, is that ia64 has two stacks, and the stack copying code that implements call-with-current-continuation needs special magic to deal appropriately. It's like porting a compiler. It's not legitimate to just assume that it will always work on any new arch. :) My belief is that we have absolutely no real evidence of guile or gnucash *ever* working on ia64. Rob, can you chime in if this is not correct? Thomas -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
