Dan Sugalski wrote:
But... this gets us very much closer to where we want to be, and I'm figuring that we're better off applying this and working out the kinks than not. I'll leave this one to Leo to make final decision on, though.
Thanks Dan for this easter egg :)
Intermediate results:
- patch applied fine
Good! (I tested it enough, so I'm happy that was successful.)
- compiles currently modulo some C++ style variable declarations
I'm lately being spoiled by C99 (which allows mid-block declarations too), but these were just oversights on my part.
and s/#import/#include/
Oops. (It's the ObjC in me.)
- a few warnings about shadowed vars
Hmm, I'll need to figure out how to turn on warnings about this in gcc. I would expect some "unused" warning as well.
- patched source does *not* conform to pdd07
Not too surprising--I'm sure there are some tabs and too-long-lines and such, that I know I need to fix.
- additional coffee breaks during compiles will ruin me finally
Yes, no kidding! At least, the makefile as currently constructed won't re-build ICU as a result of a re-Configure-ing parrot. (And it might make sense to remove cleaning of ICU from the default clean target, so that you only rebuild it if you've cleaned it explicitly.) Also, be glad I'm stopping it from building the data into a library--that takes an additional lifetime to build.
- compiled ok
Yeah!
- make test - not too bad (5/1495 subtests failed)
Let me know which ones--I'm curious as to why.
- JIted NCI (uses string_make for 't' sig) is b0rken
Ah yes, it make sense I wouldn't have hit this, if it's in the i386-specific code.
- one native_pbc test fails (more are skipped now)
I know there needs to be added some endianness correction of the serialized string data now, which hasn't been added yet, so I wouldn't be surprised if the ppc-generated file fails the test on i386, so this might be that. There are SKIPs around the other platform test files (probably should be TODOs), which will need to be regenerated on those platforms.
- make testj fails some more (8/1432 subtests failed)
I just confirmed that for me all is passing under make testj (except t/pmc/object-meths.t:17 of course), but it's not surprising for this to be platform-specific.
Thanks much for the feedback!
JEff
