This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

[RFA:] fix PR40249 crtstuff build breakage caused by inline change. Documentation needs fixing!

I won't fix the documentation; here's just a patch to fix the
build breakage.

This patch replaces -fno-inline-functions with -fno-inline when
building crtstuff.  Apparently, as Richi mentioned in the PR
trail, the recent inline changes made -fno-inline-function be
just a parameter-tweaking option, or it already was, and the
change made it just pass a threshold, so that the crtstuff
trickery with sections behind gcc's back finally broke.

Anyhow, recent-or-not, the disarmament of -fno-inline-function
is inconsistent with the documentation, which says that
-fno-inline-functions is the option to use to avoid "spontaneous
inlining".  The documentation doesn't even mention -finline at

So, hoping that the behavior is actually correct, here's a patch
in that direction to fix the build.  Now cris-elf builds again,
regtest shows no regressions since last successful build, FWIW.

Ok to commit?

	PR middle-end/40249
	* (CRTSTUFF_CFLAGS): Replace -fno-inline-functions
	with -fno-inline.

Index: gcc/
--- gcc/	(revision 147868)
+++ gcc/	(working copy)
@@ -599,7 +599,7 @@ TARGET_LIBGCC2_CFLAGS =
 # Options to use when compiling crtbegin/end.
-  -finhibit-size-directive -fno-inline-functions -fno-exceptions \
+  -finhibit-size-directive -fno-inline -fno-exceptions \
   -fno-zero-initialized-in-bss -fno-toplevel-reorder -fno-tree-vectorize \

brgds, H-P

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]