[Pkg-julia-devel] julia_1.0.0-1_amd64.changes REJECTED

Lumin cdluminate at gmail.com
Wed Aug 15 10:48:55 BST 2018


On Tue, Aug 14, 2018 at 06:00:10PM +0000, Bastian Blank wrote:
> 
> The package name does not match the library SONAME:
> 
> Package is named: libjulia1, soname "1". Package only contains SONAME "1.0"
> 
> -rw-r--r-- root/root   2093360 2018-08-14 12:28 ./usr/lib/x86_64-linux-gnu/libjulia.so.1.0

This is what I expected and is NOT a mistake.

Evidence:
  1. https://github.com/JuliaLang/julia/milestones
     > Non-breaking changes and features that can be added after 1.0.

  2. Make.inc

     155 # libjulia's SONAME will follow the format libjulia.so.$(SOMAJOR). Before v1.0.0,
     156 # SOMAJOR will be a two-decimal value, e.g. libjulia.so.0.5, whereas at and beyond
     157 # v1.0.0, SOMAJOR will be simply the major version number, e.g. libjulia.so.1
     158 # The file itself will ultimately symlink to libjulia.so.$(SOMAJOR).$(SOMINOR)

Conclusion:
  Before Julia's 1.0.0 release, every MINOR revision was a breaking
  change, but not after the 1.0.0 release.

  For example,
  1. libjulia0.6 (= 0.6.x) -> libjulia0.7 (= 0.7.x) is a breaking change.
  2. libjulia1 (= 1.0.0) -> libjulia1 (= 1.1.0) is NOT a breaking change.

  According to upstream's blue print, if libjulia1 (= 1.1.0) isn't
  compatible to libjulia1 (= 1.0.0) then it will be upstream's fault.
  That means, when we change the symlink

     libjulia.so.1 (symlink) -> libjulia.so.1.0 (ELF shared object)

  into

     libjulia.so.1 (symlink) -> libjulia.so.1.1 (ELF shared object)

  nothing is expected to break.

> Please describe directly why you need debug infos.

Justification: Unable to pass unit test without debugging info.

Evidence:
  1. debian/rules

    158 # Don't strip sys.so and libjulia.so.* as suggested upstream.
    159 # https://github.com/JuliaLang/julia/issues/23115#issuecomment-320715030
    160 # 1. stripping sys.so would cause "backtrace" test failure.
    161 # 2. not stripping libjulia.so* would improve the results of the backtrace.

  2. experiment:

     (1) debomatic-amd64 / unstable : julia 1.0.0-1 : autopkgtest
	     debugging information is kept
	     -> all upstream unit test passed

     (2) debomatic-amd64 / experimental : julia 1.0.0-1 : autopkgtest
	     debugging information is stripped
		 -> unit test failure

Conclusion:
  1. keeping debugging info makes sense.

@ginggs: could you please verify whether my understanding is correct?
Apart from the issues mentioned above, I found some minor problems to fix.




More information about the Pkg-julia-devel mailing list