If you use pipe() you get two ends in the same process, then you fork and child and parent can communicate. This is how a unix shell setups up pipes and it is rather elegant.
back
1 comments
Doing the same thing on Windows, I create the pipe and get two ends in the same process. Then I'd call CreateProcess and indicate I want the pipe's handle (fd) inherited to the child, and I'd use a prearranged way to tell the child what the fd value is it should use.
Possibly the most common way to tell the child the value is by setting it as a CLI arg in CreateProcess.
Yes, and CreateProcess needs special facilities to make this possible while in the UNIX model you don't.
Which special facilities are you referring to? If it's the ability to selectively inherit handles (fds) to a new process, Linux's lack of this "special facility" is nothing to be proud of.
How do you selectively pass on fds without having a global impact on your process?
As explained above, you fork and then before exec you can use all the usual APIs to close/duplicate/... file descriptors. This has no global impact on the parent process because you are already in the child, but one still has access to all the context of the parent needed to whatever configuration you want to do, including things that will be invented in 20 years and are not even conceived today.
And as noted elsewhere, what you want is a new process with a specific fd: duplicating your current process state is a path-dependent evolution, rather than a logic step.
There have been plenty of comments here about effective workarounds, multi-process architectures to keep fork cheap, zygotes... these are very specifically working around the problem, while trying to avoid admitting it's a problem.
You are changing your argument. Let's first agree that my point that being able to use standard APIs (dup,close,pipe,...) is elegant and avoids having a lot of parameter for a process creation call.
I don't think it's an elegant design at all; I think it's a workable design given the constraint of requiring fork.
It's an achievable result with less specialised parts, but there's a reason we don't write all of our logic in NAND gates.
I'm still only inferring what you meant by CreateProcess needing special facilities; I assume you meant opt-in fd inheritance. Saving a parameter call while forcing all fds to be shared/duplicated seems penny-wise but pound-foolish.