| 1 |
On 10/06/13 00:38, Greg Turner wrote: |
| 2 |
> On Sat, Oct 5, 2013 at 4:32 PM, Greg Turner <gmt@×××××.us> wrote: |
| 3 |
>> Trust me, at the end of the day, it's a cygwin bug. Technically |
| 4 |
>> speaking, Cygwin supports bidi pipes with the same semantics as Linux, |
| 5 |
>> but the implementation is broken. IIRC, something as simple as "diff |
| 6 |
>> -u <(echo 1) <(echo 2)" in cygwin bash is enough to confirm that. |
| 7 |
> If I'm not mistaken, the bug has to do with getting the handles of |
| 8 |
> named pipes cloned during fork(). They are supposed to stay attached, |
| 9 |
> of course, but instead they insta-close on read... something like that |
| 10 |
> anyhow, it was a while ago I looked into it :) |
| 11 |
> |
| 12 |
|
| 13 |
From what I can see it's down to bi-directional read/write on a FIFO. |
| 14 |
|
| 15 |
Typically a FIFO expects a reader and a writer (it's just a filesystem |
| 16 |
PIPE). |
| 17 |
Therefore we should be using two file descriptors, one for read and one |
| 18 |
for write. Not one file descriptor and using read/write. That's the |
| 19 |
Linux'ism |
| 20 |
we're talking about here. |
| 21 |
|
| 22 |
Alan. |
Archives