Compilers and languages (Was: Help reading a 9 track tape

dwight dkelvey at
Tue Aug 3 09:04:02 CDT 2021

Recompiling Forth was always such a trivial process, there was no reason to not recompile itself using itself. It was also a good check of the output. One could compare the output and check any differences to ensure that they were intended. One could run it twice again as a check as well.
It wouldn't always ensure that it had no errors but would validate that only the intended changes were there. Some of the early Forths had bugs in some of the math operations.
I recall reading about such errors in the groups and recompiling the version I had to fix such errors.

From: cctalk <cctalk-bounces at> on behalf of Stefan Skoglund via cctalk <cctalk at>
Sent: Tuesday, August 3, 2021 1:33 AM
To: Gavin Scott <gavin at>; General Discussion: On-Topic and Off-Topic Posts <cctalk at>; Fred Cisin <cisin at>
Subject: Re: Compilers and languages (Was: Help reading a 9 track tape

mån 2021-08-02 klockan 20:00 -0500 skrev Gavin Scott via cctalk:
> Another interesting question is whether the currently shipping
> version
> of a language written in itself was compiled using the same version
> of
> itself or the previous version. I recall HP compilers generally being
> built with the previous version (at least the last time I looked
> which
> was probably in another century).

GNAT itself was written in Ada from the beginning, though the backend
is part of gcc so partly rewritten to support some Ada constructs
(which also benefitted C++).

>From the beginning I believe they used the Alsys compiler until
GNAT was able to compile itself.

More information about the cctalk mailing list