[Openmcl-devel] CCL rebuild fails for current (and old) master on OSX 10.14

[R. Matthew Emerson]

> On Dec 31, 2019, at 4:55 AM, Bruce O'Neel <bruce.oneel at pckswarms.ch> wrote:
> 
> 
> HI,
> 
> I have now submitted a feedback to Apple as 
> FB7515029.

Thanks for doing that.  I had already submitted FB7464217, but your test case is shorter.

I think that a lot of Apple people are out for the Christmas and New Year's Day holidays, so I'm hopeful that someone will look at it when the holidays are over.

Plainly CCL needs to change so that its assembly language code can be assembled with the LLVM assembler, but still, the older assembler should still work.



> 
> I guess that's what one does to submit a bug report to apple...  That's what I figure out under https://developer.apple.com/support/ <https://developer.apple.com/support/>.
> 
> I've slightly simplified the code.
> 
> cheers
> 
> bruce
> 
> 30 December 2019 21:51 "Bruce O'Neel" <bruce.oneel at pckswarms.ch> wrote:
> Hi,
> 
> I *think* both problems we see, the truncated message as well as the invalid movq instruction are really the same problems where a negative number is converted to an unsigned.
> 
> I think a minimal example is below.  All of the movq instructions should be the same, but the last fails. 
> 
> 
> /* assemble with as -Q -g -arch x86_64 bus.s */
>   
> /* fails with assembler included with Xcode 11 -- Xcode 11.1.2 at least */
> 
> /* works on Xcode 10.3 or earler */
> 
>         node_size = 8
>         dnode_size = 16
>         num_subtag_bits = 8
>         subtag_shift = num_subtag_bits
>         ntagbits = 4
>         fulltag_misc = 13
>         misc_bias = fulltag_misc
>         fulltag_nodeheader_0 = 5
>         subtag_catch_frame = (fulltag_nodeheader_0 | (2 << ntagbits))
> misc_header_offset = -fulltag_misc
> misc_header_offset_neg = fulltag_misc
> misc_data_offset = misc_header_offset+node_size /* first word of data    */
> misc_subtag_offset = misc_header_offset       /* low byte of header   */
> 
> 
> 
> 
>         .text
>         .align 2
>         .p2align 3
> 
>         .globl junk
> junk:
>         /* movq %rax,misc_data_offset(%rsi) */
>         /* should all be the same */
>         movq %rax,-5(%rsi)
>         movq %rax,-13+8(%rsi)
>         movq %rax,-13+node_size(%rsi)
>         movq %rax,-fulltag_misc+node_size(%rsi)
>         movq %rax,misc_header_offset_neg+node_size(%rsi)
>         movq %rax,-misc_header_offset_neg+node_size(%rsi)
>         movq %rax,node_size+misc_header_offset(%rsi)
>         /* fails */
>         movq %rax,misc_header_offset+node_size(%rsi)
>         ret
> 
> 
> 
> 30 December 2019 20:32 Dave Cooper <david.cooper at genworks.com> wrote:
> 
> 
> Maybe Apple just doesn’t know about it. 
> 
> Do we have a link to a set of minimal steps to reproduce the problem, which we could start trying to spread around on Twitter, @aapl?
> 
> 
> 
> 
> On Sat, Dec 28, 2019 at 12:44 PM Tim Bradshaw <tfb at tfeb.org <mailto:tfb at tfeb.org>> wrote:
> My guess is that it doesn't matter to them, because if it did it would have been fixed by now.  I also imaging that Apple are probably not going to listen to people saying it's a bug unless those people are big and noisy enough that they might care.  So probably it will not be fixed any time soon, and unless something happens to the bits of assembler that CCL relies on CCL will therefore stop being buildable on macOS in due course. This worries me.
> 
> --tim
> 
> 
> 
> On 18 Dec 2019, at 15:07, Bruce O'Neel <bruce.oneel at pckswarms.ch <mailto:bruce.oneel at pckswarms.ch>> wrote:
> 
> Of course if this doesn't matter to Apple then then it's just amusing and not useful :-(.  To me this is a clear bug and not necessary just because CCL needs the assembler to work correctly.
> 
> 
> _______________________________________________
> Openmcl-devel mailing list
> Openmcl-devel at clozure.com <mailto:Openmcl-devel at clozure.com>
> https://lists.clozure.com/mailman/listinfo/openmcl-devel <https://lists.clozure.com/mailman/listinfo/openmcl-devel>
> --
> 
> My Best,
> 
> Dave Cooper, david.cooper at gen.works
> genworks.com <http://genworks.com/>, gendl.org <http://gendl.org/>
> +1 248-330-2979
> 
> 
> <>
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.clozure.com/pipermail/openmcl-devel/attachments/20191231/f9ef4285/attachment.htm>