From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jeffrey Perry To: David Edelsohn Cc: law@cygnus.com, jperry@xionics.com, egcs@egcs.cygnus.com, jperry@xionics.com Subject: Re: egcs1.1.2 lnx cross ppc newuser followup Date: Fri, 30 Apr 1999 23:15:00 -0000 Message-ID: <9904022256.AA25737@dogbert> References: <9904022029.AA41906@marc.watson.ibm.com> X-SW-Source: 1999-04n/msg00056.html Message-ID: <19990430231500.GmaFIh9u8neen66BKMZZ2WCHuQ_5n_3taVmFLoe-yLM@z> Hmm... I know there are linker scripts now installed in /usr/local/ppc-rtems-eabi/libs /ldscripts (binutils 2.9.1 installed them) I really don't know anything about them. I did poke around and find the symbols in target formatted object files. ppc-rtems-eabi-nm ecrtn.o | grep -i sdata 00000000 G __SDATA2_END__ 00000000 G __SDATA_END__ ppc-rtems-eabi-nm ecrti.o | grep -i sdata 00000000 G __SDATA2_START__ 00000000 G __SDATA_START__ Now what? Why couldn't it find these, they were in gcc/ > >>>>> Jeffrey A Law writes: > > Jeff> 1. Is is possible these symbols are supposed to be defiend by the linker > Jeff> via a linker script? I believe this is the case. > > Jeff> 2. Or are they supposed to be provided by the libc on your target system? > Jeff> Less likely than #1, but possible. > > I know that GCC/EGCS explicitly provides those symbols in assembly > file "crt" bookends for eABI. I thought that Linux provided its own > crti/crtn bookends for the link step. The LinuxPPC configuration seems to > specifically look for crt1.o, crti.o, crtbegin.o, crtend.o, crtn.o. It > even looks for ecrti.o and ecrtn.o if -mnewlib is used. > > David >