| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Limitations that need addressing still:
- Doesn't forcibly terminate children that should have exited but haven't
Enable with env var ENABLE_ZYGOTE_MANAGER=1.
BUG=11841
TEST=
start the browser, then make chrome and all .pak files unreadable; or alternately, start an installed browser, and uninstall the browser while it's running. Then create a new tab and browse to two new sites.
Here's an example script to hide and unhide the .pak files (note: do not move the directory they're in, that doesn't work):
#!/bin/sh
chmod_all() {
chmod $1 sconsbuild/Debug/chrome
for path in . locales obj/chrome/app/intermediate/repack obj/global_intermediate/* themes
do
chmod $1 sconsbuild/Debug/$path/*.pak
done
}
case $1 in
hide) chmod_all 000 ;;
show) chmod_all 755 ;;
esac
Review URL: http://codereview.chromium.org/115773
git-svn-id: svn://svn.chromium.org/chrome/trunk/src@17840 0039d316-1c4b-4281-b951-d872f2087c98
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
On POSIX systems, system calls can be interrupted by signals. In this
case, they'll return EINTR, indicating that the system call needs to
be restarted.
(The situation is a little more complicated than this with SA_RESTART,
but you can read man 7 signal if you like.)
The short of it is that you need to catch EINTR and restart the call
for these system calls:
* read, readv, write, writev, ioctl
* open() when dealing with a fifo
* wait*
* Anything socket based (send*, recv*, connect, accept etc)
* flock and lock control with fcntl
* mq_ functions which can block
* futex
* sem_wait (and timed wait)
* pause, sigsuspend, sigtimedwait, sigwaitinfo
* poll, epoll_wait, select and 'p' versions of the same
* msgrcv, msgsnd, semop, semtimedop
* close (although, on Linux, EINTR won't happen here)
* any sleep functions (careful, you need to handle this are restart
with different arguments)
We've been a little sloppy with this until now. This patch adds a
macro for dealing with this and corrects every case of these system
calls (that I found).
The macro is HANDLE_EINTR in base/eintr_wrapper.h. It's safe to
include on Windows and is a no-op there.
On POSIX, it uses GCC magic to return the correct type based on the
expression and restarts the system call if it throws EINTR.
And you can use it like:
HANDLE_EINTR(close(fd));
Or:
ssize_t bytes_read = HANDLE_EINTR(read(fd, buffer, len));
*BEWARE* that it will evaluate the argument multiple times, so this is
not safe:
HANDLE_EINTR(close(FireMissiles()));
http://groups.google.com/group/chromium-dev/browse_thread/thread/41a35b2a457d73a0
http://codereview.chromium.org/100225
git-svn-id: svn://svn.chromium.org/chrome/trunk/src@15102 0039d316-1c4b-4281-b951-d872f2087c98
|
|
|
When forking a child process, one often wants to move existing file
descriptors to well known locations (stdout, stderr etc). However,
this is a process bedeviled with corner cases. Consider the case where
you open two file descriptors, get assigned fds 1 and 0 for them and
wish to shuffle them so that they end up in slots 0 and 1. Our current
code fails in this case.
We also have a problem where we're currently trying to mark file
descriptors as close-on-exec rather than closing them in the child
process. This is inherently broken in a multithreaded process where
another thread can open a file descriptor and race the loop which is
trying to mark them.
Thus, on Linux we switch to close-after-fork where we known that no
other threads exist in the process. On Mac, the code is sufficiently
different that this simple fix isn't applicable and one of the Mac
folks will need to take a look.
http://codereview.chromium.org/100127
BUG=11174
git-svn-id: svn://svn.chromium.org/chrome/trunk/src@14978 0039d316-1c4b-4281-b951-d872f2087c98
|
