Date: Mon, 11 Mar 2013 07:58:22 +0200 From: Konstantin Belousov <kostikbel@gmail.com> To: Ilya Kaliman <ilya.kaliman@gmail.com> Cc: freebsd-hackers@freebsd.org Subject: Re: Linking produces unusable executable Message-ID: <20130311055822.GK3794@kib.kiev.ua> In-Reply-To: <CA%2BSmX2BCY=ySp-1aXHRHWa2LbCb31vBGvu4ivwyzAi2M4S_ujQ@mail.gmail.com> References: <CA%2BSmX2BCY=ySp-1aXHRHWa2LbCb31vBGvu4ivwyzAi2M4S_ujQ@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
--lKueCj4EYK2W7OYQ Content-Type: text/plain; charset=us-ascii Content-Disposition: inline Content-Transfer-Encoding: quoted-printable On Sun, Mar 10, 2013 at 09:00:02PM -0400, Ilya Kaliman wrote: > Hello, hackers! >=20 > I have a strange problem with a big executable. I have a piece of scienti= fic > software and some C++ module in it with a LOT of template code. When comp= iled > this module produces 450 MB archive file (w/o debugging symbols). Now, wh= en I > compile the program without this module everything works perfect. With th= is > module turned on the linker produces an executable (around 180 MB in size) > without any errors or warnings. But when I try to start the final program= zsh > just says: abort. ldd on this exe says: ./a.out: signal 6. >=20 > I watched the memory consumption during linking and it doesn't seem to ex= haust > all available memory (the linker seem to stop allocating on around 2 GB).= I've > also tried to enable --no-keep-memory for ld with no luck - linking still > produces no errors but the resulting executable is unusable. >=20 > I've tried it on 9.1 and 10-CURRENT with both gcc/g++/ld from the base sy= stem > and from ports (gcc 4.7.3, binutils 2.23.1) and with clang. >=20 > I've tried to build some of the bigger ports like chromium (just in case): > all works fine. >=20 > Everything works on linux though (with the same gcc/ld). With debugging s= ymbols > the exe is around 1GB, without them its around 200MB. Works fine in every= case > with different optimization levels. >=20 > Any ideas how to solve this? For start, it would be nice to provide some useful information together or even instead of the long story. What is the architecture ? Show at least the output of the size(1) on the final binary. Show exact shell message on the attempt of the binary run. Show the ktrace/kdump of the start attempt. As a guess, look at the sysctl kern.maxtsiz and compare it to the text size reported by the size(1). The same for the initialized data segment and kern.maxdsiz. --lKueCj4EYK2W7OYQ Content-Type: application/pgp-signature -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.19 (FreeBSD) iQIcBAEBAgAGBQJRPXJ9AAoJEJDCuSvBvK1BqbYP/2g3sQxERXQQ21Sxuvey6lpF 2Cp1KERyXmHyUoPmC9DPXykarsGffS8+z8138/0go2gYr84w1d++WVgdwsT8soAk eXbbxKv76/TlVD5P27GLX0v7m4H1KLQYOrr1bVKS2F7Qt5pfj0k9xLch557nHGnk 5a2w9PN/61WP4Lvjx+8m/hU7Tp/vwzy0Ol91hPgHHVrw/VCGO1sTW3GWaHDcg1Q9 uNs6wiBY97Oes/y563feK5MG+BlKCBsuTNZTb0nGxKlBEYxU3sRDJYfxVAYyyUUC 4svtbLt3iyvvlxC4uWbxuzR/YZWX0mYnwWJVpqRH46d3d0Q8SODV1N0Wcnzq0YJV 7FYywUGTUGFnlFYXnZ7JgJb659VdIoKHkSZ3fc00gLPCSzL9hU5FNKSvCdYrV37j wpm1x5hmg12yTnEEDh4/g1CXTRQsqCEKSSd2CsnttMnInym9WbXNN4h56y5R1sY7 o3LRCY2egB35yWY/TSHsDh3eOEMIRjB9Rc/2AIiAfcOdAJNRgOxwFF0NATC7/QKg /Si5/HyUawKqiRl9SbQyXTlS4AJr8YCp388H+XWZLYR0xfG6aQ5X1CnWLEKZoing v8FstUkXjfWhCtF7qixrA1TTXfizc3kOozDbcOa1ZnT52zQwzNzRVxT3i6GyCIGV JLML4A17D2vID2yxRkPK =Sxze -----END PGP SIGNATURE----- --lKueCj4EYK2W7OYQ--
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20130311055822.GK3794>