Infrastructure: allow passing stdin to subprocesses. #1134
Conversation
|
Maybe we should support a streaming interface for #1043 so that it is not necessary to put the whole file chunk into RAM. |
| void close_handle_mark_invalid(HANDLE& target) noexcept | ||
| { | ||
| auto to_close = std::exchange(target, INVALID_HANDLE_VALUE); | ||
| if (to_close != INVALID_HANDLE_VALUE && to_close) |
There was a problem hiding this comment.
Why is this checking to_close for null here -- and should that be a fatal error since it's an invariant violation?
There was a problem hiding this comment.
Some APIs use 0 as 'invalid handle', others use INVALID_HANDLE_VALUE. The intent is for this to work with either without having to worry about which is which.
It seems like 1043 should work by attaching the file directly rather than needing to describe the right concurrency ... thing that would have to go here. |
BillyONeal
force-pushed
the
process-stdin
branch
from
72f7336 to
e109bb6
Compare
…eads-stdin.c. Co-authored-by: Robert Schumacher <[email protected]>
BillyONeal
force-pushed
the
process-stdin
branch
from
e109bb6 to
3439c70
Compare
… pipe in order to resolve deadlocking trying to write stdin while stdout is pending.
…n moves for everything. Always explicitly name all handles to inherit when launching children. RAIIize control C handling.
BillyONeal
force-pushed
the
process-stdin
branch
from
d7ac556 to
c78bf7f
Compare
Co-authored-by: Kevin Lalumiere <[email protected]>
klalumiere
left a comment
klalumiere
left a comment
There was a problem hiding this comment.
Thanks for the attribution for the changes in downloads.cpp! 🙂
5 participants
We need this capability to call
git mktree.The temporary-file-less fix to #1144 was authored by https://github.com/klalumiere