> On Sep 28, 2017, at 4:04 PM, Russell Senior <russ...@personaltelco.net> wrote: > >>>>>> "Philip" == Philip Prindeville <philipp_s...@redfish-solutions.com> >>>>>> writes: > > Philip> Hi. I’m trying to update Perl from 5.24.1 to 5.26.1 but running > Philip> into some issues. > > Philip> We don’t use ./Configure to build the target versions (just the > Philip> host version), so when new settings are added, we need to figure > Philip> out what they are… and what the appropriate settings are for all > Philip> processors. > > Philip> And word is from upstream that 5.28.1 will have even more stuff > Philip> that we need to add to get it to build. > > Philip> Wondering if there isn’t an easier way to figure out > Philip> automatically what those settings should be. > > Philip> Obviously we can’t compile and run for other platforms, but we > Philip> can compile and extract information from those images (with > Philip> objdump, nm, etc). > > Philip> A lot of the tests that get run during Configure (grep ‘$run > Philip> ./try’ Configure) are run just to dump out information about the > Philip> result of compilation… most of which could also be extracted > Philip> just by examining the generated object (such as with objdump > Philip> -d). > > Philip> Anyone have any suggestions for things we can try to make > Philip> ./Configure work for cross-compiles as well? > > Philip> Because doing version updates seems to involve a fair amount of > Philip> guesswork about what the correct values are for a whole gamut of > Philip> platforms, and this seems error-prone. > > A long time ago, I tracked down a bug in the perl builds on a brcm47xx > device by installing a native toolchain in an nfs-mounted rootfs and, > over about 24 hours, running the auto configuration script on the target > device. > > I have no idea if the situation has improved since then, maybe 10 years > ago. >
That sounds about as fun as spinning up a Qemu in the makefile and running configure inside that… -Philip _______________________________________________ Lede-dev mailing list Lede-dev@lists.infradead.org http://lists.infradead.org/mailman/listinfo/lede-dev
