Re: 3.5 package/binary size curiosity
Hi,
On Fri, Apr 06, 2012 at 12:05:34PM +0200, Rene Engelhard wrote:
> We don't use "special" options: I only can explain it with gcc 4.6 producing
> bigger stuff than whatever ancient the official builds use (did 4.6 also be
> bigger than upstream builds), but I more think that upstream might just build
> with -Os.
FWIW, Yes, that is the case:
see http://patch-tracker.debian.org/patch/series/view/libreoffice/1:3.5.1-1/debian-opt.diff:
diff --git a/solenv/inc/unxlngi.mk b/solenv/inc/unxlngi.mk
index d86aab1..a10adb6 100644
--- libreoffice-3.5.0/solenv/inc/unxlngi.mk
+++ libreoffice-3.5.0/solenv/inc/unxlngi.mk
@@ -26,7 +26,6 @@
#*************************************************************************
# mk file for Unix Linux Intel (X86) using GCC, please make generic modifications to unxlng.mk
-CDEFAULTOPT=-Os
.IF "$(GCCNUMVER)" >= "000400050000"
ARCH_FLAGS*=-mtune=atom
.ELSE
So assuming you talked about i386 - yes, upstream builds with -Os and we
with -O2 and that would explain us being bigger
(in contrast to that amd64 is built on both per -O2)
Regards,
Rene
Reply to: