#71 Result of AC_FUNC_MALLOC is ignored

[Anonymous]

The result of the AC_FUNC_MALLOC test is discarded. For a negative result malloc is redefined to rpl_malloc which in turn is not provided by psmisc.

[Craig Small]

It's not, but rpl_malloc is universally ignored from what I see. A better idea might be to just remove the test. However, is there some setup you have that actually requires a replacement malloc?

[Anonymous]

[DarKRaveR]

I wanted to add some more info, here'S the output from configure with tests that fail and make the build bail out later on:
checking for struct user_regs_struct.orig_eax... no
checking for struct user_regs_struct.eax... no
checking for struct user_regs_struct.ebx... no
checking for struct user_regs_struct.ecx... no
checking for struct user_regs_struct.edx... no
checking for struct user_regs_struct.orig_rax... no
checking for struct user_regs_struct.rax... no
checking for struct user_regs_struct.rdi... no
checking for struct user_regs_struct.rsi... no
checking for struct user_regs_struct.rdx... no
checking for struct pt_regs.orig_gpr3... no
checking for struct pt_regs.gpr... no
checking for struct pt_regs.uregs... no
checking for struct pt_regs.regs... no
checking for struct pt_regs.cp0_status... no
checking for struct pt_regs.orig_d0... no
checking for struct pt_regs.d0... no

and:

checking for GNU libc compatible malloc... no
checking for working memcmp... no
checking for stdlib.h... (cached) yes
checking for GNU libc compatible realloc... no

Which will lead to:

peekfd.c:252:29: error: ‘struct user_regs_struct’ has no member named ‘REG_ORIG_ACCUM’
(and quite some more)

and

killall.c:(.text+0x44): undefined reference to `rpl_malloc' (and dozens more)

I managed to build psmisc by setting
ac_cv_func_malloc_0_nonnull=yes
ac_cv_func_realloc_0_nonnull=yes

I don't know if this is safe though, and it does not feel like fixing the actual reason.

And I need to add -DX86_64 to CFLAGS, as a fast nasty workaround for the first problem.