I don't see any problem with it, particularly since a0 has already been clobbered. In addition, return the proper error response in a0 will be dectected by jemalloc, which will fall back to using sbrk() and begin allocating memory out of the assigned data segment.That's great! So we don't have to move USRSTACK anymore? Or should we do it, now since we are at it? :)
I wouldn't mind being able to use up to 256 megs on my A1200 during bulk builds.
Nice. Ok, I have already compiled a full 5.99.18 release which I will install now and make some regression tests.
Were the fixes committed? Or did you patch locally? John