From mboxrd@z Thu Jan  1 00:00:00 1970
Content-Type: multipart/mixed; boundary="===============0299650374404391726=="
MIME-Version: 1.0
From: Josh Stone <jistone@redhat.com>
To: elfutils-devel@lists.fedorahosted.org
Subject: Re: [PATCH] libdw: Correct spelling of DW_LANG_PLI in dwarf.h.
Date: Wed, 26 Oct 2016 11:18:53 -0700
Message-ID: <8762b06a-ee1f-e79f-93d2-94cc80eae11b@redhat.com>
In-Reply-To: 20161026174745.GD10081@stream

--===============0299650374404391726==
Content-Type: text/plain; charset="utf-8"
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable

On 10/26/2016 10:47 AM, Mark Wielaard wrote:
> On Wed, Oct 26, 2016 at 09:36:31AM -0700, Josh Stone wrote:
>> On 10/26/2016 03:24 AM, Mark Wielaard wrote:
>>> The name used in the standard and other DWARF implementations for
>>> Programming Language One, PL/I, is DW_LANG_PLI (not DW_LANG_PL1).
>>> [...]
>>> +Version 0.168
>>> +
>>> +libdw: dwarf.h corrected the DW_LANG_PLI constant name (was DW_LANG_PL=
1).
>>> +       Any existing sources using the old name will have to be updated.
>>> +
>>
>> That typo has been there since the dawn of elfutils.git, 2005-07-26
>> commit b08d5a8fb42f.  It seems callous to break API here, even if it was
>> wrong.  Why not include a #define to preserve compatibility?
> =

> Since it doesn't break ABI we have historically just fixed such wrong
> constant names, see e.g. the removal in 0.160 of the non-existing
> DW_TAG_mutable_type in 0.160. Given that others (libdwarf, binutils)
> do define the correct name in their headers.

It doesn't matter what other headers do -- if someone was already using
the bad constant in elfutils' headers, that's an API break.  But sure,
you might still decide that's acceptable.

> Also I would be somewhat afraid that might break the generation
> of known-dwarf.h (although I didn't try).

It might be ok for known-dwarf.h to know both PL1 and PLI.

> Is there actually any source out there that uses this (wrong)
> constant name?

https://github.com/search?q=3Ddw_lang_pl1&type=3DCode

Lots of hits, but in the first few pages they're all just clones of
dwarf_aggregate_size.c.

via Google I found at least one real hit from an elfutils user:
https://lists.gforge.inria.fr/pipermail/simgrid-commits/2015-September/0222=
87.html

and it seems some version of eclipse cdt had the same typo:
https://www.cct.lsu.edu/~rguidry/eclipse-doc36/src-html/org/eclipse/cdt/uti=
ls/debug/dwarf/DwarfConstants.html

and also found the typo in a dwarfstd issue:
http://www.dwarfstd.org/ShowIssue.php?issue=3D130819.1


>>> -    DW_LANG_PL1 =3D 0x000f,	     /* ISO PL/1:1976 */
>>> +    DW_LANG_PLI =3D 0x000f,	     /* ISO PL/1:1976 */
>>
>> You should correct the comment too, "PL/I".
> =

> The '1' in the comment really is correct, because that
> is the actual name of the standard. That is probably where
> the confusion came from.

Figure 8 in the DWARF 3 and 4 standards both call it "ANSI PL/I:1976"
(not "ISO"), as does Table 3.1 in the DWARF 5 draft I've got.

Wikipedia also uses "PL/I", and even has a redirect from "PL/1".

This "I" is pronounced "one" though, surely as a roman numeral.

--===============0299650374404391726==--