Hi Michael,
I'm inclined to promote the "forkables" code to master. I just have a
few points before we do that:
- Revert bumping of CYGWIN_VERSION_SHARED_DATA. We only have to do that
if the shared region changes in an incompatible way. But this is just
adding a member to the end.
- I'm looking a bit cross-eyed on the usage of forkables_needs and
cygwin_shared->prefer_forkable_hardlinks. It seems to me as if the
split between those two isn't quite right and the fact that both
share information seems error prone.
IMHO prefer_forkable_hardlinks should actually be the single marker
for the per-installation state. After startup of the first process it
should be "unknown" (0) by default. At startup it's set to one of
"disabled" (-1) no NTFS or dir is missing
"enabled" (+1) NTFS and dir exists
That sets the state once and for all by the first Cygwin process in
the system.
Consequentially, forkables_needs should only reflect the per-process
states
"needless"
"needed"
"created"
- Shouldn't forkables_needs be renamed to forkables_needed?
That's all. There are a few minor formatting issues, but they are
negligible.
Thanks,
Corinna
signature.asc
Description: PGP signature
