Path: csiph.com!news.mixmin.net!border2.nntp.ams1.giganews.com!nntp.giganews.com!buffer2.nntp.ams1.giganews.com!nntp.brightview.co.uk!news.brightview.co.uk.POSTED!not-for-mail
NNTP-Posting-Date: Sat, 12 Nov 2016 13:04:59 -0600
From: Gareth Owen <gwowen@gmail.com>
Newsgroups: comp.lang.c
Subject: Re: Compiler design in C?
References: <nvsmpq$noi$2@dont-email.me> <nvtbjd$ahk$2@dont-email.me> <nvthtd$3pu$2@dont-email.me> <87shr14n5b.fsf@gmail.com> <nvutra$30v$1@dont-email.me> <87r36k1l4w.fsf@gmail.com> <3e47c9bc-1c0a-40dc-9dc3-92e88ace6c31@googlegroups.com> <o00196$hgr$1@dont-email.me> <o006bg$5g2$1@dont-email.me> <o01ph4$rco$1@dont-email.me> <ce04217b-bcf0-4b29-8a58-9b6ffabf9688@googlegroups.com> <o02hqq$ndk$1@dont-email.me> <d5d5655d-ff29-4273-8791-1d09c6fbf385@googlegroups.com> <o02q7a$oca$1@dont-email.me> <23d0f58e-a17c-4ef7-8db6-9cc5b7bcaf04@googlegroups.com> <o040re$d5a$1@dont-email.me> <9e27a17d-aaab-48e1-9139-95114dea0da6@googlegroups.com> <o05b51$81t$1@dont-email.me> <3121b7be-aca1-44b0-9dc9-6b23611bbfba@googlegroups.com> <o074os$61m$1@dont-email.me> <300e8fda-56cf-476f-9bd7-257791357090@googlegroups.com>
Date: Sat, 12 Nov 2016 19:04:59 +0000
Message-ID: <8760nsr02s.fsf@gmail.com>
User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/24.5 (gnu/linux)
Cancel-Lock: sha1:BSAJRqtOPT2/1/DgkUEO32LbJkE=
MIME-Version: 1.0
Content-Type: text/plain
Lines: 21
X-Usenet-Provider: http://www.giganews.com
X-Trace: sv3-FIi40qJHCnhKPPTfy9Q71e/v+TfpHCSxRuhSbzKhL9wQFozAAMvZc7zftCvs1sCJDTXGzcuwpBk3Esz!U+SgAvHb78ohdFTMSH7i7faEygpXxewG0RWozpqSpQZm16IEoLzpEKtJy1a+JQUvO7ZB
X-Abuse-and-DMCA-Info: Please be sure to forward a copy of ALL headers
X-Abuse-and-DMCA-Info: Otherwise we will be unable to process your complaint properly
X-Postfilter: 1.3.40
X-Original-Bytes: 2772
Xref: csiph.com comp.lang.c:96418

supercat <flatfinger@casperkitty.com> writes:

> The Standards allow compilers to do many things.  I would suggest, however,
> that people making quality compilers should endeavor to see how they are
> being used, and provide a means of producing break-proof builds (e.g. have
> a command-line option that would prevent a compiler from enabling any new
> optimizations that might break old code, except when such options are
> expressly enabled).  

There's a compatibility mode in gcc-6 that makes it work like gcc-4.6
Simply install gcc-4.6 alongside gcc-6 and invoke it as gcc-4.6

> Do you see any disadvantage to a compiler doing those things?

If they're possible.  But optimizers don't really work like that.  For
example, two different peephole optimizers will produce different code
based on the order they're invoked.  Would changing the order of
invocation qualify as a "new optimization"?

If so, you're insisting that the internal structure of the compiler
must, essentially, freeze, and that's not going to happen.