On 11 January 2014 22:57, Andrei Gherzan <[email protected]> wrote: > > On Sat, Jan 11, 2014 at 11:45 PM, Bernhard Reutner-Fischer > <[email protected]> wrote: >> >> On 11 January 2014 21:42:00 Andrei Gherzan <[email protected]> wrote: >> >>> Hello all, >>> >>> I am trying to compile uclibc in thumb mode - which used to work in the >>> past. Here is a log: >>> http://pastebin.com/Y1hLKTvL >> >> >> Thumb only is supposed to work, yes. >> >>> >>> One of the errors seem to be related to >>> commit 0eaaf662b70806181b7fe2a1b165769f4e48d0f0 >>> Author: Bernhard Reutner-Fischer <[email protected]> >>> Date: Fri Jun 15 13:44:35 2012 +0200 >>> ldso: use .arm mode for resolver unconditionally >>> as per comment in the file. >>> Fixes runtime with __THUMB_INTERWORK__ enabled. >>> Signed-off-by: Bernhard Reutner-Fischer <[email protected]> >>> >>> Isn't this breaking thumb? I'm talking about processors which support >>> only >>> thumb - no ARM instructions support. >> >> >> Abovementioned commit was made under the assumption as per commit comment, >> but I am pretty sure that I used a thumb1 with interworking enabled so this >> could be wrong indeed. What CPU are you compiling for, which tune and insn >> encoding? > > > In the above pastebin you can see the gcc flags used and CPU which is a > Cortex M3 (-mthumb -mno-thumb-interwork -mfloat-abi=softfp -mtune=cortex-m3 > -march=armv7-m)
I think i have touched this single spot now in master to be encoding agnostic (Paul added thumb2 in efce79f09ae6daa77cd322df0d532beec3f445f5 ) but It seems to me it was/is not complete: >From the looks there are a couple of spots setting the thumb bits on addresses that are written in arm mode, so you (resp. your thumb2 only v7 cortex-m3) will trip. Patches to fix thumb2-only are very welcome, of course. _______________________________________________ uClibc mailing list [email protected] http://lists.busybox.net/mailman/listinfo/uclibc
