From mboxrd@z Thu Jan 1 00:00:00 1970 From: Kai Ruottu To: Juan Manuel Dodero Cc: insight@sourceware.cygnus.com Subject: Re: Cross debugging to erc32 with Insight Date: Wed, 09 Feb 2000 01:42:00 -0000 Message-id: <38A141FE.5872169B@freenet.hut.fi> References: X-SW-Source: 2000-q1/msg00058.html Juan Manuel Dodero wrote: > > I have downloaded two versions of Insight. The one that I can build without > problems (1999-10-04) doesn't cross-debugs to erc32. I try to configure > cross-debugging with these command lines: > > ./configure --target=sparc-erc32-rtems --host=i686-pc-linux-gnu > > gcc -c -g -O2 -W -Wall -I/install/Linux/insight-19991004/itcl/itk/unix -I/ > insc > make[2]: *** No rule to make target It seems that you have clipped the important info away, the 'insc' isn't anything sane, there should have been the real error message too... I remember the Insight 'configure' leaving some subdirs unconfigured, e.g. when the itk build needs the 'tkConfig.sh' (or something) from the 'tk/unix' subdir, the build will fail and say that it is missing. Then one just configures in the 'tk' subdir and continues the build from the main build dir, by writing 'make'... Solving these mechanical problems isn't however enough... My experiences with the ESA/ESTEC 'erc32' are somehow confused... The original binary format seems to be 'aout', later there was the 'coff' format used and now the egcs-20000131 snapshots handles the 'sparc-*-rtems*' as the same target as the 'elf'-based 'sparc-*-rtemself'... So the important question is the binary format needed for your tools? What does your GCC now produce? What format do you binutils handle? Which format they should use? The 'coff_sparc' seems to be a compromise between the 'aout' and 'elf', e.g. the 'xgc' distribution seems to use it (although a little modified) : D:\opt\xgc\bin>erc-coff-ld -V GNU ld version 2.8.1 (with BFD 2.8.1) Supported emulations: coff_erc dem32 Configuring Insight-20000131 with the target name you used, results into some weird configuration where the 'gdb/solib.c' (for shared libs in SunOS and Solaris2) will be tried to be compiled. Unfortunately it doesn't seem to succeed under Linux... And the simulator in 'sim/erc32' will not be used at all... The 'sis' included with Insight seems also to be the version 2.7.5, but the docs from the ERC32 2.0.6 release (August 1999) talk about the sis version 3.0.5. So Insight includes a little old code (perhaps this doesn't matter)... There seems to be no special support for RTEMS in the Insight sources, so the 'rtems' can be left away from the target name and replaced with the more important binary format name. One target I know to work ok is the Fujitsu 'sparclite-coff'. Using this target name, the simulator in 'sim/erc32' will be built into GDB/Insight. How the 'sparclite' differs from the 'erc32' type and if targets like 'sparc-aout', 'sparc-coff' or 'sparc-elf' (or with the '-erc32', like 'sparc-erc32-elf') could also work, I cannot say... But telling the binary format sounds more reasonable than trusting the 'rtems' to give some sane configuration. Ok, downloading the ERC32 distribution sources and docs from ESA/ESTEC ( ftp://ftp.estec.esa.nl/pub/ws/wsd/erc32/erc32ccs ) could give more info. For me it looks like the FSF and sourceware sources and the official ERC32 distribution aren't very well synchronized... Cheers, Kai