Details
-
Bug
-
Status: Open (View Workflow)
-
Minor
-
Resolution: Unresolved
-
5.5.57
-
Linux LEDE/OpenWRT SDK
Description
trying to build mariadb 5.5.x on arm/mips targets with musl it gives me the following error
*../libmysql/libmysqlclient.a(my_context.c.o): In function `my_context_spawn_internal':
|
my_context.c:(.text+0x3c): undefined reference to `setcontext'
|
my_context.c:(.text+0x40): undefined reference to `setcontext'
|
../libmysql/libmysqlclient.a(my_context.c.o): In function `my_context_continue':
|
my_context.c:(.text+0xa4): undefined reference to `swapcontext'
|
my_context.c:(.text+0xc4): undefined reference to `swapcontext'
|
../libmysql/libmysqlclient.a(my_context.c.o): In function `my_context_spawn':
|
my_context.c:(.text+0x140): undefined reference to `getcontext'
|
my_context.c:(.text+0x15c): undefined reference to `getcontext'
|
my_context.c:(.text+0x180): undefined reference to `makecontext'
|
my_context.c:(.text+0x1c0): undefined reference to `makecontext'
|
../libmysql/libmysqlclient.a(my_context.c.o): In function `my_context_yield':
|
my_context.c:(.text+0x204): undefined reference to `swapcontext'
|
my_context.c:(.text+0x218): undefined reference to `swapcontext'
|
collect2: error: ld returned 1 exit status*
|
with glibc it works properly, x86_64 and i386 are also working with musl and glibc
patching the my_context.h file like this
@@ -31,8 +31,6 @@
|
#define MY_CONTEXT_USE_X86_64_GCC_ASM
|
#elif defined(__GNUC__) && __GNUC__ >= 3 && defined(__i386__)
|
#define MY_CONTEXT_USE_I386_GCC_ASM
|
-#elif defined(HAVE_UCONTEXT_H)
|
-#define MY_CONTEXT_USE_UCONTEXT
|
#else
|
#define MY_CONTEXT_DISABLE
|
#endif |
is working, what could be the right approach ?
Attachments
Activity
cat CMakeCache.txt | grep UCONTEXT
HAVE_UCONTEXT_H:INTERNAL=1
in gcc-5.4.0/fixincludes/tests/base/sys I have
cat ucontext.h
/* DO NOT EDIT THIS FILE.
It has been auto-edited by fixincludes from:
"fixinc/tests/inc/sys/ucontext.h"
This had to be done to correct non-standard usages in the
original, manufacturer supplied header file. */
#if defined( LINUX_IA64_UCONTEXT_CHECK )
- define _SC_GR0_OFFSET \
__builtin_offsetof (struct sigcontext, sc_gr[0])
#endif /* LINUX_IA64_UCONTEXT_CHECK */
in config.h I have this
#define HAVE_UCONTEXT_H 1
–
PS: I don;t know why formatting is not working for me
Yes, so HAVE_UCONTEXT_H is set, but apparently it cannot be used. The fix is to test it in configure.cmake.
Your formatting didn't work because {{ ... }} is for inline code, for multi-line code snippets use {code} ... {code}.
so if it can't be used I think is safe to assume MY_CONTEXT_DISABLE and use the patch above in building the image without messing with configure.cmake
?
I don't know how to test for the functionality and maybe this is a musl problem and not on mariadb side
Yes. You're probably right both counts. This looks like something musl would omit.
Does it mean that you have <ucontext.h>, but it is not usable?
If yes, I'd say the correct approach is to check in CMakeLists.txt whether ucontext.h is usable. In configure.cmake there's a check:
CHECK_INCLUDE_FILE(ucontext.h HAVE_UCONTEXT_H)IF(NOT HAVE_UCONTEXT_H)CHECK_INCLUDE_FILE(sys/ucontext.h HAVE_UCONTEXT_H)ENDIF()and instead of (or together with) this it should use CHECK_C_SOURCE_COMPILES.