| Age | Commit message (Collapse) | Author |
|---|
|
|
|
|
|
Sticky shift key...
|
|
|
|
The shell's pseudo-variable is called LINES. This is true of at least
bash and mksh.
|
|
|
|
|
|
|
|
Calling unshare(2) immediately puts us in the new namespace
with the "overflow" user and group ID. By calling geteuid()
and getegid() in handle_r() after calling unshare(), we try
to map that to root, which Linux refuses to let us do.
What we really want to map to root is the caller's uid/gid
in the original namespace. So we have to save them before
calling unshare().
|
|
ioctl(). (This is a thing Android's old mount already does.)
|
|
On Android, much of the restorecon logic is in libselinux, so this
isn't portable. We do want to be able to build on the host for
testing *other* toys, though, so #if keeps this building.
Change-Id: Ida5a6713a926140c549d5770d62798f4aedca748
|
|
Also switch to xopen for O_CLOEXEC paranoia and to avoid a conditional.
Change-Id: Iee5c4c124bcac800313f586768ffcaade542bd22
|
|
Change-Id: Icbee1df8b316039c68ef326f9d4d80f32c603b21
|
|
Change-Id: I83c478ea9ddd456c61e3f0f99506a1028a7027a1
|
|
> Yes, I know, "don't use pending". Sadly, more(1) is pretty useless
> without this. It gets confused by long lines or tabs.
>
> This patch also adds the missing prompt between multiple files.
actually, it looks like we're already hard-coding some escape
sequences? more(1) doesn't need anything that isn't in ANSI, so here's
an alternative patch that fixes the same bugs as the other patch but
also has a reverse-video prompt:
|
|
Not much progress this week. Just route and umount.
Also remove a stray reference to load_policy which already switched.
Also remove an outdated claim that Android logging is done in the
kernel. It's been in userspace for a while now.
|
|
|
|
|
|
|
|
Guys, you CANNOT COPY CODE FROM BUSYBOX. I don't care if I was the initial
author of that file, other people have touched it since.
I know I added "pending" because I couldn't keep up with code review in
realtime and stuff was getting lost, so we needed a place to park things
before they underwent the full dorodango process. I also know none of the
binaries I distribute has anything in pending enabled. But dude, if I'm going
to find stuff like this I have to start reviewing the code BEFORE applying
it to pending, and we all know where that leads.
Please don't do this again.
|
|
|
|
|
|
Can't think of a better place to stick the actual version info than a #define
at the top of main. Makefile and toys.h are equally awkward, querying git
tags wouldn't work if building a release tarball... watch me forget to update
this when cutting a release.
|
|
|
|
confusing update-alternatives, the paths of the links installed by toybox should
match those installed by busybox. This is accomplished by changing the flags
of a few tools within toybox.
|
|
|
|
|
|
|
|
Checking in a stopping point where I figured out what I did wrong, before changing it.
|
|
|
|
argument of -n).
|
|
|
|
|
|
|
|
Four obsolete tools were removed from Android, and runcon moved to toybox.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
toybox "make menuconfig" currently fails to build if the compiler is in C99
mode (clang 3.6 and gcc 5.0 snapshots default to C99 mode, you can also
reproduce the failure with older gcc by using 'make menuconfig HOSTCC="gcc
-std=gnu99"').
The problem is that inline semantics changed in C99, and kconf_id_lookup's
inline-ness depends on C89 semantics there.
|
|
Note that this is a case where Android's tool isn't the same as the
usual tool. Ours takes an explicit file containing the policy to be
loaded. restorecon is at least command-line compatible, but the
implementation is all in Android's libselinux where there's a
selinux_android_restorecon function.
|
|
|
|
|
|
|
|
Also, coreutils says "s" instead of "seconds". POSIX specifies the
format of the previous two lines, but doesn't even mention this line.
|
