Re: [ECOS] Error building from CVS sources

[Randall Nortman <ecoslists at wonderclown dot com>]

On Tue, Oct 18, 2005 at 09:29:26AM -0400, Randall Nortman wrote:
[...]
> Not seeing a configure option for this (e.g., --without-gprof), I just
> deleted the gprof portion of the source tree and ran configure again.
> This time it apparently completed, so I ran make, which also, to my
> great surprise, completed successfully.  However, I'm looking for the
> "configtool" program, referenced in the documentation, but under
> host/tools/configtool, only ecosconfig is being built.  Is this
> adequate?  Has configtool disappeared since the 2.0 release?  Or am I
> doing something wrong?
> 
> Also, RedBoot is not being built.  In fact, hardly anything is being
> built.  After a "make && make install", these 12 files are the only
> things installed:
[...]
> So, how am I supposed to be doing this?  I've found no README or any
> docs for building from source, so I'm sort of flying blind here.

I have, to some extent, answered my own question, with a little more
digging.  I've managed to get ecosconfig to create a build tree for
redboot, though running "make" in this tree results in lots of syntax
errors, which I describe below.  It still might be nice to have the
graphical configtool available, if only so I can follow along with the
examples in the documentation.  (I generally prefer command-line tools
myself, but eCos seems to be a big hairy beast to a newcomer, so the
comfort of a GUI might be nice at first.)  So, where does configtool
live?  How do I get it/build it?

OK, so the syntax errors are in packages/cygmon/current/misc/arm/cpu.h.
The problem starts on line 448, at the definition of the __cli()
macro, which is inline asm code.  The problem is that this macro
definition spans multiple lines without line continuation backslashes.
So, I added backslashes to the __cli and __sti macros.  Errors then
pop up in monitor_cmd.c line 432, where there is a similar problem
with missing backslashes.  It seems that this code assumes that quoted
strings can cross lines without backslashes.  Perhaps that was true
with an older version of gcc, but my version (4.0.2) apparently thinks
otherwise.

I could keep chugging through the code adding backslashes, but I
wonder if perhaps there's something I'm missing.  Is this code known
to be incompatible with gcc 4.0.2?  Am I doing something wrong?  Is it
a bug in gcc v4, or is gcc fixing an old non-standard behavior?  (I'm
not quite sure myself what the ANSI C spec says about line breaks in
quoted strings.)  I do notice that some long strings are backslashed
in the code; just a few are not.  Perhaps these are recent commits to
CVS that just haven't been fully tested yet?

TIA,

Randall

-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss