MIME-Version: 1.0
Date: Thu, 30 Jun 2011 15:01:06 -0500
Message-ID: <BANLkTinM3VV4sxtoyzSGVqqssxrN2tzPNQ@mail.gmail.com>
Subject: BinUtils (miscellaneous)
From: Rugxulo <rugxulo AT gmail DOT com>
To: djgpp-workers AT delorie DOT com
Content-Type: text/plain; charset=ISO-8859-1
Reply-To: djgpp-workers AT delorie DOT com

Hi guys,
   (Note to DJ: I know you're probably VERY busy, and I think even an
email or two to you bounced recently, ugh. So feel free to ignore
this. And yes I see your online compiler thingy hack is offline too,
heh.)

1). Andris, what exactly are the holdups from having newer BinUtils? I
mean, I honestly don't need or want it (and the changes seem fairly
minor, esp. for us!), just vaguely curious (esp. in light of the
following).

2). CWS updated his ED (EDT clone) to 1.6 about a year or so ago. In
it he had in his TODO to make configure.com be DOS-friendly.
Eventually he just renamed it, but I got inspired and used some rough
data from FASM's forum to make a 7-bit ASCII executable. Okay, so all
it does is say, "DOS is not VMS!" and quit. But that's better than
crashing, no? (configure.com is a text file for VMS, but there is no
way for DOS or 32-bit NTVDM to know this.) Is this worth giving to
BinUtils maintainers? (Apparently DJ is one!) Doubt it, but ... I
already emailed the VMS dude (Tristan), and he seemed (very
unsurprisingly) nonchalant. So whatever, perhaps somebody here can
convince them if anybody agrees with me (doubt it).

http://sites.google.com/site/rugxulo/vmscom.asm?attredirects=0

; hY1X-00P[h:!_)?)?CC)?)?h]1X-00PZh09X-00AdAc!DOS is not VMS!$


Silly, I know, but so am I.  :-)

3). Does Martin S. still read here? I finally used his hack / tips to
make a pure .COM with GAS + LD. So yeah, after a year when he first
told me, I finally converted the dumb Befunge-93 interpreter to it. It
works ... except newer BinUtils (GAS) seem to have a problem.

E.g. I'm on PuppyLinux (Lucid 5.25) now, which uses 2.20.1, and it
doesn't work. It's too confusing to remember exactly (and not obvious
anyways). It seems to work fine with 2.16.1 but 2.20.1 or 2.21 both
don't like some uses of "LEA eax,[0]" or similar. But I can work
around that. What I *can't* fix is it saying a spurious error message
about "undefined symbol L0 not found". Only DJ himself (or somebody
smarter than I am) could possibly decipher it. It's a new "change",
and I'm not sure why!

BTW, in case you didn't guess, I'm willing to bet that BinUtils won't
accept the dumb VMS/DOS patch unless it uses their tools. (I doubt
they use FASM.) Not hard to do, but I'm assuming they don't want
hard-coded .byte constants!! Well, it's probably moot, they probably
don't care anyways!! (Though it could still bite anyone using 32-bit
Windows too, e.g. MinGW peeps.) But still ... Martin's hack doesn't
work anymore.   :-(     It's not even a linker problem but something
in GAS (regression??). Well, obviously most people don't (shouldn't?)
use GAS for this kind of stuff, heh.

4). Last I checked, --gc-sections didn't work for COFF at all.
Apparently the COFF backend maintainer is ... DJ himself! But I know
he's always busy working. So I don't pester him much. Anyways,
somebody at SourceWare (?) did add a rough patch (for 2.20) to add
such support for MS / PE COFF recently. I didn't try it, but perhaps
somebody here is more adventurous.

http://sourceware.org/bugzilla/show_bug.cgi?id=11539

(Yeah, lots of boring trivia, feel free to ignore ....)