Feedback-ID: id4a34324:Fastmail
Date: Sat, 6 Aug 2022 19:46:23 -0700
From: Andres Freund <andres@anarazel.de>
To: Thomas Munro <thomas.munro@gmail.com>
Cc: Tom Lane <tgl@sss.pgh.pa.us>, Andrew Dunstan <andrew@dunslane.net>,
	Robert Haas <robertmhaas@gmail.com>,
	pgsql-hackers <pgsql-hackers@postgresql.org>
Subject: Re: Cleaning up historical portability baggage
Message-ID: <20220807024623.v62stodeyondbpvc@awork3.anarazel.de>
References: <1249825.1658622186@sss.pgh.pa.us>
 <CA+hUKGLryBW4fNYoXzMG9y1e0H-QOTdSiTG46xCr1DNuuPwbeg@mail.gmail.com>
 <CA+hUKGJMcwbzRyfTocfxbPBUhERqKE42d-L22jQ+vpB7an-NjQ@mail.gmail.com>
 <20220804013546.h65najrzig764jar@awork3.anarazel.de>
 <3295936.1659577924@sss.pgh.pa.us>
 <20220804023010.24snkyq3hser5pxq@awork3.anarazel.de>
 <CA+hUKGLjffiULiVs5K4mi8FO5dgfngjKx5n9vZdDzQB9hzfurw@mail.gmail.com>
 <CA+hUKGJJxtpG12XFu6XJRxo+1ZmTrnYwY9_e8FgrVS7K9ar5Mg@mail.gmail.com>
 <20220805210823.u546w4p4wzfany7p@awork3.anarazel.de>
 <CA+hUKGLyGrqbC_YgW_+Fa1mLKev-WGmta1cDC5nVp+iuQ3bE0A@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: 
 <CA+hUKGLyGrqbC_YgW_+Fa1mLKev-WGmta1cDC5nVp+iuQ3bE0A@mail.gmail.com>
Archived-At: 
 <https://www.postgresql.org/message-id/20220807024623.v62stodeyondbpvc%40awork3.anarazel.de>
Precedence: bulk

Hi,

On 2022-08-07 11:47:31 +1200, Thomas Munro wrote:
> So what about strtof?  That's gotta be dead code too.  I gather we
> still need commit 72880ac1's HAVE_BUGGY_STRTOF.

> From a cursory glance at MinGW's implementation, it still has the
> complained-about behaviour, if I've understood the complaint, and if I'm
> looking at the right C runtime[1].

Well, right now we don't refuse to build against the "wrong" runtimes, so it's
hard to say whether you're looking at the right runtime. I don't think we need
this if we're (as we should imo) only using the ucrt - that's microsoft's,
which IIUC is ok?


> -/*
> - * strtof() is part of C99; this version is only for the benefit of obsolete
> - * platforms. As such, it is known to return incorrect values for edge cases,
> - * which have to be allowed for in variant files for regression test results
> - * for any such platform.
> - */

We can't remove the result files referenced here yet, due to the double
rounding behaviour?

Greetings,

Andres Freund