On 24.04.2013 20:28, KP Kirchdoerfer wrote: > Am 24.04.2013 19:00, schrieb Andrew: >> 24.04.2013 19:10, KP Kirchdoerfer пишет: >>> Hi Andrew; >>> >>> I like the idea, but the toolchain is AFAIK related to the architecture >>> (see uClibc config), so we will still need one toolchain for each >>> architecture, am I wrong? >> Right, toolchain is dependent from architecture (for ex., ARMv5, ARMv6 >> and so on), and not from platform (SoC - TI OMAP, Broadcom, Samsung >> Exynos and so on). If all SoCs are ARMv6 - they can use single >> toolchain. And single userspace apps. But for each platform (couple of >> similar SoCs from same vendor that can boot same kernel - for ex., >> Ralink RT3050/3052/3350/3352/5350/5352 and maybe RT3662/3883) we need >> separate kernel with modules. > > Just to get it right - I think uClibc config is the main point we > declare our toolchain; it offers i386, x86_64, arm and a lot more, but > not ARMv5, ARMv6.... > But I see that our make/toolchain/ files does probably not distinguish > between uclibc's architecture and the kernel... > > ? > kp > Right, uClibc config + gcc target arch. I was wrong about armv5/v6, this is specified like i486/i686 make options. Toolchain files are dependent of target architecture (gcc is built for specified target arch), and are dependent from kernel headers (that are generated at building 'linux' package).
------------------------------------------------------------------------------ Try New Relic Now & We'll Send You this Cool Shirt New Relic is the only SaaS-based application performance monitoring service that delivers powerful full stack analytics. Optimize and monitor your browser, app, & servers with just a few lines of code. Try New Relic and get this awesome Nerd Life shirt! http://p.sf.net/sfu/newrelic_d2d_apr _______________________________________________ leaf-devel mailing list leaf-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/leaf-devel