220 39736 article Path: news.gmane.org!.POSTED!not-for-mail From: Niall Douglas Newsgroups: gmane.comp.lang.c++.isocpp.proposals Subject: Re: D1095R0/N2xxx draft 4: Zero overhead deterministic failure - A unified mechanism for C and C++ Date: Sun, 12 Aug 2018 15:12:40 -0700 (PDT) Lines: 151 Approved: news@gmane.org Message-ID: References: <206ed6c7-67da-43db-8d21-5d2a359b19b0@isocpp.org> <05e6c694-3efe-4caa-8557-0edd07bc0c5f@isocpp.org> Reply-To: std-proposals@isocpp.org NNTP-Posting-Host: blaine.gmane.org Mime-Version: 1.0 Content-Type: multipart/mixed; boundary="----=_Part_1238_129812676.1534111960377" X-Trace: blaine.gmane.org 1534111838 30362 195.159.176.226 (12 Aug 2018 22:10:38 GMT) X-Complaints-To: usenet@blaine.gmane.org NNTP-Posting-Date: Sun, 12 Aug 2018 22:10:38 +0000 (UTC) To: ISO C++ Standard - Future Proposals Original-X-From: std-proposals+bncBDGKFT5YZADRBWPBYLNQKGQEYKTBFEI@isocpp.org Mon Aug 13 00:10:34 2018 Return-path: Envelope-to: gclcip-std-proposals@m.gmane.org Original-Received: from mail-yw1-f71.google.com ([209.85.161.71]) by blaine.gmane.org with esmtp (Exim 4.84_2) (envelope-from ) id 1foyZE-0007nE-2k for gclcip-std-proposals@m.gmane.org; Mon, 13 Aug 2018 00:10:32 +0200 Original-Received: by mail-yw1-f71.google.com with SMTP id p8-v6sf20689933ywl.14 for ; Sun, 12 Aug 2018 15:12:42 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=isocpp-org.20150623.gappssmtp.com; s=20150623; h=date:from:to:message-id:in-reply-to:references:subject:mime-version :x-original-sender:reply-to:precedence:mailing-list:list-id :list-post:list-help:list-archive:list-subscribe:list-unsubscribe; bh=3Ios/z0Fopgh3btMkOsKJ2K8tJgM51jMiE8t8NK7K/Y=; b=TWQZTeZhhVpyqeU94pzBJ01rZFrpLAGBuv4ibklABPFA2XBTF0SDx9E32Sm5kFzvEj bXSXbMt/aHluVZNSl2iEM1oXC6gc01qMpQCBG0JhNXqZa6ndU8eV6Da0gNLaMpWyyqiO vuH+oDvOBd6S0/LHJnsl8PxmlZt1G+zio21QsnUE9CoWg/QqNzhxaiSKnhXvRQ6xxkq+ WWLXtQuli5kqHVPRYEpYsTZnBi7dZ26DlSZ0R1kWjVcCWh45UONkgjnar3KPepk1Slqi yqDrt2x7yHIjy6SDWHfP4L/vGdVPaTiMX8xtfZQYB6AL2dw94C2dvXhkNupv+uG3NGW0 3OQg== DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20161025; h=date:from:to:message-id:in-reply-to:references:subject:mime-version :x-original-sender:reply-to:precedence:mailing-list:list-id :list-post:list-help:list-archive:list-subscribe:list-unsubscribe; bh=3Ios/z0Fopgh3btMkOsKJ2K8tJgM51jMiE8t8NK7K/Y=; b=U3rHQbgHuJvaDoFPGWeFPddoiHNtYuAcnrP+nRXUFtx7QDs0yyxBAxSpyJ9CsBfdE8 bFQoFvqjfyDaO++DithifeVyx+vUU+fGm1L3JV4m09/qHnbPhu5pJGEG+dBBrwqg1TFw 04izHUB647RLUBcGs2iS8Dg1XildXAmBPz4S8YyrJzbETxzgBdTpxlEm6iUmnZVnf81d ZETwwInjggT1vHswuOxGmDS1D/PFxgx8MV+DcA2V1VDYGDj/t/RwKX6+XkN1PG+JSae6 t8dbejZzAfgUwQl9ZymIvhZfYucE9IpxG+TKpki9+K/nRNcLxIALBNdORJNUNNfnIPy3 yiIg== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:date:from:to:message-id:in-reply-to:references :subject:mime-version:x-original-sender:reply-to:precedence :mailing-list:list-id:x-spam-checked-in-group:list-post:list-help :list-archive:list-subscribe:list-unsubscribe; bh=3Ios/z0Fopgh3btMkOsKJ2K8tJgM51jMiE8t8NK7K/Y=; b=F0vYCUSq7YTyHOLjCpT/IFUmZhNzLjf4MzNdnEbvi6zAUIbIoIzO1Nk3C4RpIpXtad YsK0Iav+LmL8P9EmpQS2ZR1FSn2jY/84XsV3Tm5NxZBbZGl1InmNLvci5M8grVZ5VRzc lAR8PpnkCU9ei4uIXK98U3Imzd7toTHC3XLtPFO/ladWnzvzVufRn0nC1iqEgxDX1Q1S P/aVI1WGl+UVy6RBcZeIfrSVjZ1GElLzHztKqiBEZRyj9FYtsFtT7Uw63ocqAqmSgEAx jsBSfMNZVV0536/TpGdWmdaup3nOJtZMuJGXQwA0kVuQjWBYUOxktxaL+Upbsk9u4837 hGEA== X-Gm-Message-State: AOUpUlHVSd0dxedkCjyjqhwxFn3zXYQ99KIqjItfgvqgb4aBWa/h5DuL AF5o2GgxLceVDZYPbItLpIM6og== X-Google-Smtp-Source: AA+uWPxQ8Ra+eqzl7CCAeQ4stxh4MNRYUxvEIaWGiDvphygUIsG6hjC79EUKKAT8GEjSL5JLlxooxg== X-Received: by 2002:a81:98cb:: with SMTP id p194-v6mr4664473ywg.9.1534111962290; Sun, 12 Aug 2018 15:12:42 -0700 (PDT) X-BeenThere: std-proposals@isocpp.org Original-Received: by 2002:a81:b1c8:: with SMTP id p191-v6ls2421284ywh.3.gmail; Sun, 12 Aug 2018 15:12:41 -0700 (PDT) X-Received: by 2002:a81:5705:: with SMTP id l5-v6mr318313ywb.6.1534111960865; Sun, 12 Aug 2018 15:12:40 -0700 (PDT) In-Reply-To: <05e6c694-3efe-4caa-8557-0edd07bc0c5f@isocpp.org> X-Original-Sender: nialldouglas14@gmail.com Precedence: list Mailing-list: list std-proposals@isocpp.org; contact std-proposals+owners@isocpp.org List-ID: X-Spam-Checked-In-Group: std-proposals@isocpp.org X-Google-Group-Id: 399137483710 List-Post: , List-Help: , List-Archive: List-Subscribe: , List-Unsubscribe: , Xref: news.gmane.org gmane.comp.lang.c++.isocpp.proposals:39736 Archived-At: ------=_Part_1238_129812676.1534111960377 Content-Type: multipart/alternative; boundary="----=_Part_1239_1512024551.1534111960377" ------=_Part_1239_1512024551.1534111960377 Content-Type: text/plain; charset="UTF-8" > > > 1. The names left and right are not descriptive for their uses. I would > prefer expected and unexpected or value and error or something like that. > Already changed to _EitherValue() and _EitherError(). > b) The proposed rewrite from a setting of errno + a return statement to > a return _Failure() seems scary as the programmer may insert some other > code between the errno setting and the return statement. > Optimisation already significantly rewrites the programmer's code. Everything is "as if" nowadays. > c) The code that shows the call site for the abs() example is not > equivalent to the original as the function may set errno to 0, which is > taken as no error by the original code but is still not "positive" by the > new means. > Upon further inspection it may be that this feature is just not > explained clearly enough. I think your intention is that within a > _Fails(errno) marked function errno is a local variable. If it is non-zero > when a return statement is hit the transformation occurs. > This seems to work better than my initial understanding that it is > the code pattern of setting errno followed by return that is transformed. > Correct, real errno becomes exclusively the compiler's responsibility when calling a _FailsWithErrno function. > d) It is unclear to me what happens if errno is already != 0 when abs() > is called. To get backwards compatibility it seems logical that the local > errno gets set from the threadlocal errno when abs is entered. If this is > the case, please show how the expression: > abs(a) + abs(b) is translated (with the problem being that if > abs(a) lazily returns an error then abs(b)'s initing of its errno would use > a stale value). > If you declare your function _FailsWithErrno, it needs to be ported to the new system. Specifically, that you can't read errno. Trying to do so fails the compile. For code which calls a _FailsWithErrno function, we know such functions never read errno, but they as-if may set it. The compiler optimises accordingly. > > 3. In your last C example it is unclear if the code is the user written > code or the compiler rewrite, and in the latter case if the user had a > errno = 0 statement at all. > I will clarify the wording and spell out why and how statements were elided by the compiler due to having no possible side effect. Niall -- You received this message because you are subscribed to the Google Groups "ISO C++ Standard - Future Proposals" group. To unsubscribe from this group and stop receiving emails from it, send an email to std-proposals+unsubscribe@isocpp.org. To post to this group, send email to std-proposals@isocpp.org. To view this discussion on the web visit https://groups.google.com/a/isocpp.org/d/msgid/std-proposals/eba9269f-99e8-4ed5-86ad-c05512e56f9e%40isocpp.org. ------=_Part_1239_1512024551.1534111960377 Content-Type: text/html; charset="UTF-8" Content-Transfer-Encoding: quoted-printable

1. The names left and right are not descriptive for= their uses. I would prefer expected and unexpected or value and error or s= omething like that.

Already cha= nged to _EitherValue() and _EitherError().
=C2=A0
=C2=A0 =C2=A0b) Th= e proposed rewrite from a setting of errno + a return statement to a return= _Failure() seems scary as the programmer may insert some other code betwee= n the errno setting and the return statement.
<= div>
Optimisation already significantly rewrites the programm= er's code. Everything is "as if" nowadays.
=C2=A0
=C2=A0 =C2=A0c) The code that shows the call site for the abs() exam= ple is not equivalent to the original as the function may set errno to 0, w= hich is taken as no error by the original code but is still not "posit= ive" by the new means.
=C2=A0 =C2=A0 =C2=A0 =C2=A0Upon furth= er inspection it may be that this feature is just not explained clearly eno= ugh. I think your intention is that within a _Fails(errno) marked function = errno is a local variable. If it is non-zero when a return statement is hit= the transformation occurs.
=C2=A0 =C2=A0 =C2=A0 =C2=A0This seems= to work better than my initial understanding that it is the code pattern o= f setting errno followed by return that is transformed.

Correct, real errno becomes exclusively the compil= er's responsibility when calling a _FailsWithErrno function.
= =C2=A0
<= div>=C2=A0 =C2=A0d) It is unclear to me what happens if errno is already != =3D 0 when abs() is called. To get backwards compatibility it seems logical= that the local errno gets set from the threadlocal errno when abs is enter= ed. If this is the case, please show how the expression:
=C2=A0 = =C2=A0 =C2=A0 =C2=A0abs(a) + abs(b) is translated (with the problem being t= hat if abs(a) lazily returns an error then abs(b)'s initing of its errn= o would use a stale value).

If = you declare your function _FailsWithErrno, it needs to be ported to the new= system. Specifically, that you can't read errno. Trying to do so fails= the compile.

For code which calls a _FailsWithErr= no function, we know such functions never read errno, but they as-if may se= t it. The compiler optimises accordingly.
=C2=A0

3. I= n your last C example it is unclear if the code is the user written code or= the compiler rewrite, and in the latter case if the user had a errno =3D 0= statement at all.

I will clari= fy the wording and spell out why and how statements were elided by the comp= iler due to having no possible side effect.

Niall= =C2=A0

--
You received this message because you are subscribed to the Google Groups &= quot;ISO C++ Standard - Future Proposals" group.
To unsubscribe from this group and stop receiving emails from it, send an e= mail to std-proposa= ls+unsubscribe@isocpp.org.
To post to this group, send email to std-proposals@isocpp.org.
To view this discussion on the web visit https://groups.google.= com/a/isocpp.org/d/msgid/std-proposals/eba9269f-99e8-4ed5-86ad-c05512e56f9e= %40isocpp.org.
------=_Part_1239_1512024551.1534111960377-- ------=_Part_1238_129812676.1534111960377-- .