在 2017年4月1日星期六 +08 上午10:20:28，Коля Гурьев 写道： > 31.03.2017 15:51, Boyuan Yang пишет: > > It seems that upstream code would give up when compiling if the host is > > not > > x86 or x86_64. However, such information and defined macros were not used > > anywhere in the source code. Thus I think we could simply patch away the > > blocking #error macro. > > One of these macros is used in Telegram/SourceFiles/ui/animation.h:169 > > #ifdef ARCH_CPU_32_BITS > #define SHIFTED_USE_32BIT > #endif // ARCH_CPU_32_BITS > > #ifdef SHIFTED_USE_32BIT
Sorry for overlooking it. This should be the sole place using that macro. That means building on 64bit little endian architectures is supposed to work theoretically out of box with previous patch. (ppc64el and arm64). Adding extra checks may support armhf as well. Mips64el might work as well (but launchpad lacks such architecture support). Anyway we could always limit its building architectures by writing them explicitly in debian/control if wanted. (e.g. see virtualbox or wine) > ... > > Also have a look at Telegram/SourceFiles/config.h:276 > > #if Q_BYTE_ORDER == Q_BIG_ENDIAN > #error "Only little endian is supported!" > #endif // Q_BYTE_ORDER == Q_BIG_ENDIAN I saw it. Maybe we shouldn't build on big endian machines. -- Sincerely, Boyuan Yang
signature.asc
Description: This is a digitally signed message part.
