Please read this before reporting a bug:
https://wiki.archlinux.org/title/Bug_reporting_guidelines
Do NOT report bugs when a package is just outdated, or it is in the AUR. Use the 'flag out of date' link on the package page, or the Mailing List.
REPEAT: Do NOT report bugs for outdated packages!
https://wiki.archlinux.org/title/Bug_reporting_guidelines
Do NOT report bugs when a package is just outdated, or it is in the AUR. Use the 'flag out of date' link on the package page, or the Mailing List.
REPEAT: Do NOT report bugs for outdated packages!
FS#61517 - [julia] 2:1.0.3-2 code_native broken
Attached to Project:
Community Packages
Opened by Jakob Nybo Nissen (jakobnissen) - Thursday, 24 January 2019, 14:05 GMT
Last edited by Alexander F. Rødseth (xyproto) - Tuesday, 26 March 2019, 13:35 GMT
Opened by Jakob Nybo Nissen (jakobnissen) - Thursday, 24 January 2019, 14:05 GMT
Last edited by Alexander F. Rødseth (xyproto) - Tuesday, 26 March 2019, 13:35 GMT
|
DetailsDescription:
The code_native function does not work properly on the Arch Linux binaries. This seems to be an issue specific to the binaries package, and does not occur when using the official binaries. The issue is described in more detail in the thread: https://github.com/JuliaLang/julia/issues/28046 - note that this issue covers two distinct issues that can occur during building of Julia, and only the behaviour under the heading "Fedora Behavior" is seen with this package (not the behaviour seen on MacOS). Additional info: Package version: 2:1.0.3-2 Steps to reproduce: 1) Start Julia in an interactive REPL. 2) Execute "@code_native 1+1", without the quotes. 3) This will erroneously print "WARNING: Could not determine size of symbol" instead of the expected output. |
This task depends upon
Closed by Alexander F. Rødseth (xyproto)
Tuesday, 26 March 2019, 13:35 GMT
Reason for closing: Upstream
Tuesday, 26 March 2019, 13:35 GMT
Reason for closing: Upstream
They also have a fix, which was merged to master a month ago but not in the version they just released two days ago. Wonderful. I guess they don't think it matters very much.
Oh! By the way, the thing they "fixed" has never been an issue in the system llvm. One more reason to dislike their bundled private llvm code copy...
disasm.cpp:811:80: error: no matching function for call to ‘llvm::Target::createAsmStreamer(llvm::MCContext&, std::remove_reference<std::unique_ptr<llvm::formatted_raw_ostream>&>::type, bool, bool, llvm::MCInstPrinter*&, llvm::MCCodeEmitter*&, llvm::MCAsmBackend*&, bool) const’
IP, CE, MAB, /*ShowInst*/ false));
See also https://bugs.archlinux.org/task/57387
I think this issue will have to be fixed + released by upstream.