llvm dev - Jan 2015 - [LLVMdev] Should I commit "IR: Move MDLocation into place" before or after 3.6 branch?

On Wed, Jan 14, 2015 at 11:04 AM, Adrian Prantl <aprantl at apple.com>
wrote:
> One small request: Omitting the column field from MDLocation if it is 0 is
> fine, because it won’t make it into the line table. Omitting the line if it
> is 0 is not, because it gives the wrong impression that the line is being
> ignored, which is not the case. Line 0 will be emitted in the line table
> and has special semantics (Line 0 in DWARF may be used to mark
> compiler-generated code that has no corresponding source line number).
>
I don't know - I think I might be OK with it being implicitly zero (I think
that's the column behavior - if the column is never set it's just zero
in
the line table? (which is "no specific column") - that seems
consistent
with zero line ("no actual line in the source code"))

>
> thanks,
> adrian
>
> > On Jan 13, 2015, at 9:46 PM, Duncan P. N. Exon Smith <
> dexonsmith at apple.com> wrote:
> >
> > I just finished the work for PR21433.  I think it's relatively low
risk,
> > but it does change assembly output.
> >
> > Should I commit this before or after the (imminent) 3.6 branch?
> >
> > (I would have committed it tonight, but the buildbots are fairly red,
> > and I prefer to break out-of-tree code in the morning anyway.)
> >
> > <llvm.patch><clang.patch>
>
>
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20150114/c381e9f1/attachment.html>

> On Jan 14, 2015, at 11:12 AM, David Blaikie <dblaikie at gmail.com>
wrote:
> 
> 
> 
> On Wed, Jan 14, 2015 at 11:04 AM, Adrian Prantl <aprantl at apple.com
<mailto:aprantl at apple.com>> wrote:
> One small request: Omitting the column field from MDLocation if it is 0 is
fine, because it won’t make it into the line table. Omitting the line if it is 0
is not, because it gives the wrong impression that the line is being ignored,
which is not the case. Line 0 will be emitted in the line table and has special
semantics (Line 0 in DWARF may be used to mark compiler-generated code that has
no corresponding source line number).
> 
> I don't know - I think I might be OK with it being implicitly zero (I
think that's the column behavior - if the column is never set it's just
zero in the line table? (which is "no specific column") - that seems
consistent with zero line ("no actual line in the source code”))
The difference is that column 0 won’t be emitted into the line table, but line 0
will be — which is why I think it would less misleading to mirror this behavior
in the IR.

-- adrian
>  
> 
> thanks,
> adrian
> 
> > On Jan 13, 2015, at 9:46 PM, Duncan P. N. Exon Smith <dexonsmith at
apple.com <mailto:dexonsmith at apple.com>> wrote:
> >
> > I just finished the work for PR21433.  I think it's relatively low
risk,
> > but it does change assembly output.
> >
> > Should I commit this before or after the (imminent) 3.6 branch?
> >
> > (I would have committed it tonight, but the buildbots are fairly red,
> > and I prefer to break out-of-tree code in the morning anyway.)
> >
> > <llvm.patch><clang.patch>
> 
> 
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu <mailto:LLVMdev at cs.uiuc.edu>        
http://llvm.cs.uiuc.edu <http://llvm.cs.uiuc.edu/>
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
<http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev>
> 
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20150114/29d0fe25/attachment.html>

Original question is kind of moot at this point, looks like Hans has already
created the branches.

Regarding line 0, as long as "has no associated line" doesn't get
mixed up with "inherits line from last previously set line" I'm
okay.
--paulr

From: llvmdev-bounces at cs.uiuc.edu [mailto:llvmdev-bounces at cs.uiuc.edu] On
Behalf Of David Blaikie
Sent: Wednesday, January 14, 2015 11:13 AM
To: Adrian Prantl
Cc: LLVM Developers Mailing List
Subject: Re: [LLVMdev] Should I commit "IR: Move MDLocation into
place" before or after 3.6 branch?



On Wed, Jan 14, 2015 at 11:04 AM, Adrian Prantl <aprantl at
apple.com<mailto:aprantl at apple.com>> wrote:
One small request: Omitting the column field from MDLocation if it is 0 is fine,
because it won’t make it into the line table. Omitting the line if it is 0 is
not, because it gives the wrong impression that the line is being ignored, which
is not the case. Line 0 will be emitted in the line table and has special
semantics (Line 0 in DWARF may be used to mark compiler-generated code that has
no corresponding source line number).

I don't know - I think I might be OK with it being implicitly zero (I think
that's the column behavior - if the column is never set it's just zero
in the line table? (which is "no specific column") - that seems
consistent with zero line ("no actual line in the source code"))


thanks,
adrian
> On Jan 13, 2015, at 9:46 PM, Duncan P. N. Exon Smith <dexonsmith at
apple.com<mailto:dexonsmith at apple.com>> wrote:
>
> I just finished the work for PR21433.  I think it's relatively low
risk,
> but it does change assembly output.
>
> Should I commit this before or after the (imminent) 3.6 branch?
>
> (I would have committed it tonight, but the buildbots are fairly red,
> and I prefer to break out-of-tree code in the morning anyway.)
>
> <llvm.patch><clang.patch>

_______________________________________________
LLVM Developers mailing list
LLVMdev at cs.uiuc.edu<mailto:LLVMdev at cs.uiuc.edu>        
http://llvm.cs.uiuc.edu
http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev

-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20150114/af70dd07/attachment.html>

> On 2015-Jan-14, at 11:26, Adrian Prantl <aprantl at apple.com> wrote:
> 
> 
>> On Jan 14, 2015, at 11:12 AM, David Blaikie <dblaikie at
gmail.com> wrote:
>> 
>> 
>> 
>> On Wed, Jan 14, 2015 at 11:04 AM, Adrian Prantl <aprantl at
apple.com> wrote:
>> One small request: Omitting the column field from MDLocation if it is 0
is fine, because it won’t make it into the line table. Omitting the line if it
is 0 is not, because it gives the wrong impression that the line is being
ignored, which is not the case. Line 0 will be emitted in the line table and has
special semantics (Line 0 in DWARF may be used to mark compiler-generated code
that has no corresponding source line number).
>> 
>> I don't know - I think I might be OK with it being implicitly zero
(I think that's the column behavior - if the column is never set it's
just zero in the line table? (which is "no specific column") - that
seems consistent with zero line ("no actual line in the source code”))
> 
> The difference is that column 0 won’t be emitted into the line table, but
line 0 will be — which is why I think it would less misleading to mirror this
behavior in the IR.
What about this:

  - Still have `MDLocation(scope: !0)` imply `line: 0`.
  - Change printout to include `line:` even when it's `0`.

There's a sane default, so we shouldn't have to specify it.
However, the value is important and relevant, so the assembly
writer *does* print it out.

Thoughts?

(Frankly I still prefer `line: 0` being omitted, but it did
seem kind of funny when I was updating tests that CHECK for
`line: 0` that they now check for `MDLocation(scope:`, so I
guess on balance I'm indifferent here.)