<Wanda[cis]>
bleh how do I make ld.so shut up about mismatched ELFCLASS32/64
<Wanda[cis]>
...
<Wanda[cis]>
thank you element
<Wanda[cis]>
(yes of course icecube is 32-bit executables. managed by a 64-bit tcl process.)
<q3k[cis]>
psychic damage from opening glibc sources and seeing GNU indentation style
<Wanda[cis]>
oh no
<Wanda[cis]>
don't worry
<Wanda[cis]>
GNU indentation style is far from the worst thing in glibc
<q3k[cis]>
anyway you should be just one nop/jmp patch away from getting rid of this warning
<q3k[cis]>
or yknow, add another preload to nop out _dl_signal_error
<Wanda[cis]>
nah I'm just going to move LD_PRELOAD setting deeper down in the stack
<Wanda[cis]>
I mean, if this turns out to actually work
<q3k[cis]>
right
<Wanda[cis]>
q3k[cis]: I don't think you can override ld.so stuff with LD_PRELOAD
<Wanda[cis]>
it does seem to work.
<Wanda[cis]>
I love toolchains.
<Wanda[cis]>
sigh. or rather.
<Wanda[cis]>
it fixed one of three suspicious issues I was trying to track down.
<Wanda[cis]>
but not the other two
<Wanda[cis]>
luckily, it fixed the most problematic one.
<Wanda[cis]>
(I'm investigating one random crash bug, one semi-common bitstream corruption bug on UL1K only, and one rare ... either bitstream corruption bug or I'm missing something subtle); it fixed the second one, allowing me to actually proceed with UL1K. first one doesn't bother me much because my test code just ignores toolchain crashes and proceeds to the next sample. that just leaves occasional rerun needed because of #3.
<whitequark[cis]>
<q3k[cis]> "enterprise_grade_malloc.so" <- just call it a "hardening strategy"
<Wanda[cis]>
let's add an infinite loop in sbtplacer to the collection