[Gcl-devel] gcl FTCBFS: uses AC_RUN_IFELSE a lot

[Camm Maguire]

Greetings, and thanks for your report!

I would love to get gcl in cross-compiling shape, but as you note, even
leaving one AC_RUN_IFELSE in place renders the whole process moot.
Taking a quick glance, the most critical test appears to be GCL's
configure time verification that it can unrandomize sbrk(). As far as I
can tell this test at least would require running compiled binaries, so
the only question remaining is whether we could do without it. This
test has served as a very useful domino break for countless segfaults
appearing mysteriously later as images are saved and reexecuted. If it
is to be retained, the only alternative I can envision is a runtime
startup test for every GCL image, but it is not clear to me if this
could ever reliably be implemented, as each executable has a fixed,
though random, sbrk() breakpoint even when the personality flags are
broken.

I will try to think about this a bit more.

Take care,