orion@nuchat (Roland Dunkerley) (01/13/89)
I think I have found a disagreement between gcc 1.32 and gas "1.28 (I
guess.)" (that's the version number I get back from gas -v). Anyway,
gcc produces lines of the form .ln ### (where ### is presumably a line
number in the source file) in the .s file produced by gcc-cc1, and
these are not supported by gas. Any suggestions? Also on another
note, although I will make a formal announcement later when I package
up my fixes to ship to rms, I've discovered a problem with robotussin
and the /lib/crt1.o on our SysV/386 box, it seems that robotussin
doesn't know how to handle .init sections properly. My fix was to
edit the file /lib/crt1.o (actually a copy of it so as not to destroy
the original) replacing .init with .text, I then ran the resultant
file through ld as follows:
ld oocrt1.o crt1.ld -o ocrt1.o
where crt1.ld contains:
SECTIONS
{
.text: { }
.data: { }
.bss: { }
}
-----
This causes the two .text sections in oocrt1.o to be consolidated into
one .text section. the resulting file runs through robotussin fine.
Note that I have not tested this yet, due to the aforementioned
problem with gas. More on this later when I know more. Oh, by the
way, many thanks to those who worked on the gas and gdb 386 ports, gdb
inside emacs is a BIG win over sdb, if only it ran the program in a
term-mode window seperate from gdb... Maybe I'll add that, what do
y'all think? (would other people like that addition too?)
more later,
Roland Pleasant Dunkerley III, KSC
*** (orion@nuchat.UUCP) (uunet.UU.NET!nuchat!orion)
*** South Coast Computing Services
*** We service Publicly Redistributable software - reasonable rates
*** Inquire within