Dear Wolfgang: Am 27.10.14 19:59 schrieb(en) Wolfgang Denk:
I have to admit that I am not in a position yet to give any recommendations based on deeper knowledge or understanding; I cannot even tell if the observed effects are known GCC bugs, or if the fact that the same effects are no longer present in 4.9.1 is because a bug was detected and fixed, or just by pure chance.
O.k., got your point...
THis applies at least to U-Boot and Kernel, or more generaly to all kinds of code where you take special efforts when arranging things like the stack layout and/or the image linking.
I see...
I can only say that I currently use gcc 4.7.x (from ELDK 5.4) for all production builds where I need working code, resp. gcc 4.9.1 (what will become ELDK v5.7) for testing. This is just my $ 0.02. Your mileage may vary.
Which is actually *really* helpful, thanks a lot! I used ELDK 4.2 happily for years, and was forced to move when the kernel starting with 3.2.50-something failed to compile with it. My 5200 still u-boots with a /very/ old 4.2 build (never saw a reason why I should re-compile perfectly working code). For me, kernel and user space /seem/ to work with the 5.5.2 builds. I only saw some strange effects in user space after the update (delayed or lost signals between processes), though, which (IIRC) were not visible in the 4.2 builds, and which I blamed to me being too dumb (which is probably the case anyway). I think your approach is perfectly right, so I'll downgrade to 5.4... Thanks again, Albrecht.
pgpYLrmUTg68X.pgp
Description: PGP signature
_______________________________________________ eldk mailing list [email protected] http://lists.denx.de/mailman/listinfo/eldk
