On 13 January 2014 20:56, Alon Zakai <[email protected]> wrote: > Oh ok, it is meant to be run just through llc, then it fails. > > Looks like the issue is that the pnacl passes leave a type in there, > > %union.anon = type { float } > > So the problem occurred before this step, and the emcc-0 file would be > useful. >
I see the problem. It's that ExpandStructRegs doesn't handle nested struct types. vatest.c at https://gist.github.com/xxuejie/8384185 reproduces the bug. I'll prepare a fix. Thanks for the bug report! Mark > On Mon, Jan 13, 2014 at 8:44 PM, Alon Zakai <[email protected]> wrote: > >> That bitcode file is only 2k in size and compiles ok in fastcomp, perhaps >> something got mixed up? >> >> - Alon >> >> >> >> On Mon, Jan 13, 2014 at 6:20 AM, Xuejie "Rafael" Xiao >> <[email protected]>wrote: >> >>> Definitely! Here's the file: https://copy.com/xGKJsUSI93Fm. >>> >>> Let me know if you need more information here. Thanks for the help! >>> >>> Best Regards, >>> >>> 肖雪洁 >>> Xuejie "Rafael" Xiao >>> >>> >>> >>> On Mon, Jan 13, 2014 at 12:09 PM, Alon Zakai <[email protected]>wrote: >>> >>>> Sure, but any chance you can send me the bitcode that the backend fails >>>> on? EMCC_DEBUG=1, and grab emcc-1-linktime.bc from /tmp/emscripten_temp. >>>> >>>> - Alon >>>> >>>> >>>> >>>> On Sun, Jan 12, 2014 at 5:54 PM, Xuejie "Rafael" Xiao < >>>> [email protected]> wrote: >>>> >>>>> Thanks Mark! Now we don't have the LLVM error as before. >>>>> >>>>> However, I think this time we are hitting something in emscripten. The >>>>> same gist is updated: https://gist.github.com/xxuejie/8384185. This >>>>> time we are getting an error in JSBackend. So Alon, can you help take a >>>>> look at this? >>>>> >>>>> Thanks all of you for your help! >>>>> >>>>> Best Regards, >>>>> >>>>> 肖雪洁 >>>>> Xuejie "Rafael" Xiao >>>>> >>>>> >>>>> >>>>> On Sun, Jan 12, 2014 at 11:05 PM, Mark Seaborn >>>>> <[email protected]>wrote: >>>>> >>>>>> On 12 January 2014 04:50, Xuejie "Rafael" Xiao <[email protected]>wrote: >>>>>> >>>>>>> Mark: Thanks a lot for the patch! >>>>>>> >>>>>>> I did some test using the patch. Clang would indeed stop issuing >>>>>>> errors, however, we are still hitting some errors in llvm. A detailed >>>>>>> log >>>>>>> containing outputs before and after the patch can be found at here: >>>>>>> https://gist.github.com/xxuejie/8384185. >>>>>>> >>>>>>> So is this because of some restrictions in emscripten-fastcomp? >>>>>>> >>>>>> >>>>>> Good point. I didn't notice that because I didn't test it >>>>>> end-to-end. Here's a fix for the IR simplification passes in the LLVM >>>>>> branch: https://codereview.chromium.org/135953002 >>>>>> >>>>>> Cheers, >>>>>> Mark >>>>>> >>>>>> On Sat, Jan 11, 2014 at 11:39 PM, Mark Seaborn <[email protected] >>>>>>> > wrote: >>>>>>> >>>>>>>> On 8 January 2014 15:56, Alon Zakai <[email protected]> wrote: >>>>>>>> >>>>>>>>> If you have a simple solution in mruby source, that would be best >>>>>>>>> for you I think. If not, then we should see about fixing it in clang. >>>>>>>>> My >>>>>>>>> guess is that it is not a deep problem, just a todo, so likely not >>>>>>>>> hard for >>>>>>>>> us to fix. I've not worked on the clang code before though. >>>>>>>>> >>>>>>>>> Mark, do you know if pnacl has plans to fix that issue (2381)? >>>>>>>>> >>>>>>>> >>>>>>>> Since you asked, I've prepared a fix and posted it for review: >>>>>>>> http://lists.cs.uiuc.edu/pipermail/cfe-commits/Week-of-Mon-20140106/096991.html >>>>>>>> >>>>>>>> Cheers, >>>>>>>> Mark >>>>>>>> >>>>>>>> -- >>>>>>>> You received this message because you are subscribed to the Google >>>>>>>> Groups "emscripten-discuss" group. >>>>>>>> To unsubscribe from this group and stop receiving emails from it, >>>>>>>> send an email to [email protected]. >>>>>>>> For more options, visit https://groups.google.com/groups/opt_out. >>>>>>>> >>>>>>> >>>>>>> -- >>>>>>> You received this message because you are subscribed to the Google >>>>>>> Groups "emscripten-discuss" group. >>>>>>> To unsubscribe from this group and stop receiving emails from it, >>>>>>> send an email to [email protected]. >>>>>>> For more options, visit https://groups.google.com/groups/opt_out. >>>>>>> >>>>>> >>>>>> -- >>>>>> You received this message because you are subscribed to the Google >>>>>> Groups "emscripten-discuss" group. >>>>>> To unsubscribe from this group and stop receiving emails from it, >>>>>> send an email to [email protected]. >>>>>> For more options, visit https://groups.google.com/groups/opt_out. >>>>>> >>>>> >>>>> -- >>>>> You received this message because you are subscribed to the Google >>>>> Groups "emscripten-discuss" group. >>>>> To unsubscribe from this group and stop receiving emails from it, send >>>>> an email to [email protected]. >>>>> For more options, visit https://groups.google.com/groups/opt_out. >>>>> >>>> >>>> -- >>>> You received this message because you are subscribed to the Google >>>> Groups "emscripten-discuss" group. >>>> To unsubscribe from this group and stop receiving emails from it, send >>>> an email to [email protected]. >>>> For more options, visit https://groups.google.com/groups/opt_out. >>>> >>> >>> -- >>> You received this message because you are subscribed to the Google >>> Groups "emscripten-discuss" group. >>> To unsubscribe from this group and stop receiving emails from it, send >>> an email to [email protected]. >>> For more options, visit https://groups.google.com/groups/opt_out. >>> >> >> > -- > You received this message because you are subscribed to the Google Groups > "emscripten-discuss" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > For more options, visit https://groups.google.com/groups/opt_out. > -- You received this message because you are subscribed to the Google Groups "emscripten-discuss" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/groups/opt_out.
