[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: MiNT 1.10 re-sync



  I have encountered the following problems with this big unified patch:
  
  1) Psigintr() patches missing.
  
  2) Reading from serial devices (ie /dev/modem1) blocks forever due to
  a lacking call to checkbttys() in sleep().

I'll resend these patches; should they be relative to virgin 1.10 or to
the currently patched source?
  
  2) Howards sticky text is generally a good idea, but has one big draw-
  back. Only after few activity memory gets *very* fragmented -- I can't
  run gcc then! The only way to cure this is to reboot the system :(
  I'm not sure whether this is a bug in the implementation or the concept;
  I guess the last one is true.

Well, I must admit I didn't spend a whole lot of thought on the implementation.
How about if we change the way memory is allocated... Grab memory from one end,
except start at the opposite end for shared text regions. E.g., always start
from the top of memory when allocating in the default case - stack, bss, data,
non-shared text, and always allocate shared text regions from the bottom
of memory. The problem would still exist, but it would be lessened, I think.
As a side effect, the fast-load bit would be pretty much superfluous.

Comments? Problems?
  -- Howard