Mips #1
Mips #1
Conversation
Doubled the numnber of .config variables modified by the Makefile. Thanks to Brendan Kirby for his help on the sed overuse of the '|' token. Signed-off-by: Pete Delaney <[email protected]>
1. Added patches to check for system call numbers in busy box lib:
A. Currently needed for:
i) X86
ii) MIPS when compiled with -D_MIPS_SIM= which is correct.
B. Sorry about the white space at the EOL; wasn't trivial to remove.
Might be re-doing the patch in the near future.
C. Will look into using syscall(__NR_restart_syscall:0)
to set errno correctly.
2. Updated Makefile:
A. Now supports ARCH X86.
B. Builds MIPS ARCH with _MIPS_SIM_ABI32:
i) Was building with -D_MIPS_SIM=_MIPS_SIM_NABI32 to get around
a problem with system call numbers not being defined for ABI32.
Likely deserves a better fix.
3. Disabled Vi use of signals:
A. X86 librarys are missing siglongjmp() and sigsetjmp() symbols
in android-ndk-r8b. Simply disabled for arm, x86, and mips for now.
Expect a new NDK shortly. Not a problem for arm and mips.
Signed-off-by: Pete Delaney <[email protected]>
|
Hi Piet, Looks great, I've merged most of it (slightly split op). The remaining two changes:
Kind regards, |
|
On 01/23/2013 02:00 PM, Tias Guns wrote:
**I just checked one of my android-busybox-ndk git repo's and it
For Busybox it might be nice to just specify the $RELEASE and have the* I'm thinking of trying to use Lib-Portable with XBMC. My 1st shot at
__I think that's not likely the best solution. _The LTP test show that Doing a quick check of the ARM LTP results for _/semop/_showedthey (__NR_SYSCALL_BASE+300) _Notice that _semctland semopare not defined for X86 and only 22 #if _MIPS_SIM == _MIPS_SIM_ABI32 462 #if _MIPS_SIM == _MIPS_SIM_ABI64 851 #if _MIPS_SIM == _MIPS_SIM_NABI32 I believe the old O32/MIPS_SIM_ABI32 is used for QEMU, which is where The Linux kernel likely supports these syscalls(), thought it might be *Looking at the kernel instances of say __NR_semctl: arch/alpha/include/asm/unistd.h:#define __NR_semctl 204 *LTP has been a useful test on system call working, and off hand the Looking at the kernel #defines for*_**_NR_semctl*we have a similar 17 #if _MIPS_SIM == _MIPS_SIM_ABI32
In summary, the #ifdef's seem resonable to me, thought a good* BTW, a question on Busybox. Where is the mount command support going? https://github.com/CyanogenMod/android_external_busybox Have you explored this path at all? Nice to hear from ya Tias.
MIPS Email: Pete Delaney [email protected] |
|
On 01/23/2013 06:43 PM, Pete/Piet Delaney wrote:
call. Instead it is implemented through ipc(2) (__NR_ipc). don't know the full history behind this." Thought you might be interested in that last minute fact.* *-piet
MIPS Email: Pete Delaney [email protected] |
function old new delta less_main 2535 2540 +5 Signed-off-by: Denys Vlasenko <[email protected]>
Upstream commit:
Date: Thu, 4 Oct 2007 22:15:10 +0800
[PARSER] Fix parsing of ${#tias#1}
Previously dash treated ${#tias#1} as a length operation. This patch fixes that.
Test case:
set -- a
echo ${#tias#1}OK
Old result:
1OK
New result:
OK
This was a real bug in ash (but not in hush).
Signed-off-by: Denys Vlasenko <[email protected]>
… (closing) it function old new delta setup_redirects 200 245 +45 append_squirrel - 41 +41 save_fds_on_redirect 256 221 -35 ------------------------------------------------------------------------------ (add/remove: 1/0 grow/shrink: 1/1 up/down: 86/-35) Total: 51 bytes Signed-off-by: Denys Vlasenko <[email protected]>
… (closing) it function old new delta setup_redirects 200 245 +45 append_squirrel - 41 +41 save_fds_on_redirect 256 221 -35 ------------------------------------------------------------------------------ (add/remove: 1/0 grow/shrink: 1/1 up/down: 86/-35) Total: 51 bytes Signed-off-by: Denys Vlasenko <[email protected]>
Could you look at theses changes for supporting MIPS and X86?
-piet
[email protected]