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

[Emilio Pozuelo Monfort]

Hi,

On 21/11/2018 14:56, Graham Inggs wrote:
> Hi Bastian
> 
> My apologies in advance for doing this, but another month has passed.
> Another ping from me.
> 
> On 2018/10/25 12:24, Ian Jackson wrote:
>> Ian Jackson writes ("Re: julia_1.0.0-1_amd64.changes REJECTED"):
>>> Lumin writes ("Re: julia_1.0.0-1_amd64.changes REJECTED"):
>>>> 1. Isn't "incomplete backtrace" a sensible reason to keep debug symbols?
>>>>     Policy said "should" but not "must". Please tell me what I can do in
>>>>     order to help improve the src:julia package to satisfy the requirements?
>>>
>>> My main concern here is this: AFAICT this package has been REJECTed
>>> solely for this reason.  Why is this bug[1] a reason for a REJECT ?
>>> ISTM that it should be filed in the BTS and handled like a normal bug.
>>
>> Ping, ftpmaster ?
>>
>> Ian.
>>
>>> [1] Assuming it is a bug.  The discussion here suggests to me that it
>>> is, but it is really unhelpful to be having it on debian-devel in the
>>> context of an ftpmaster REJECTion.
>>
> 
> From the original REJECTion email from mid-August [1], there were two issues,
> but I believe both have been explained in the follow-up emails and subsequent
> uploads.

Since you are cc'ing d-devel@, I'll chime in.

I'm not sure I understand the first concern. The package name is libjulia1,
which would indeed normally point to a libjulia.so.1 library. You ship this:

-rw-r--r-- root/root  36143904 2018-08-16 07:48
./usr/lib/x86_64-linux-gnu/libjulia.so.1.0
lrwxrwxrwx root/root         0 2018-10-13 06:16
./usr/lib/x86_64-linux-gnu/libjulia.so.1 -> libjulia.so.1.0

That would be standard for a libjulia.so.1 SONAME (though one would need to look
at the SONAME header in the ELF file to know for sure). Also I don't see a
lintian warning for 1.0.0-1 (or your latest upload), so it seems alright to me.
Maybe I'm missing something, or perhaps this was rejected and reuploaded and
that's why I don't see why that comment was made.

(I'm not sure why you need all those other symlinks in
./usr/lib/x86_64-linux-gnu/julia/ to files that are not shipped anywhere. That
smells like a bug, but I haven't investigated)

As for the other issue. It is very unusual to ship a non-stripped library. In
your lintian override, you point to [2], which mentions that it's ok to ship the
debug info in a separate file (the stripped symbol files found in -dbg or
-dbgsym packages) as long as the original files have a .gnu_debuglink section
pointing to that separate file. That is what dh_strip will do for you, as you
can see in its manpage and by inspecting the binary with readelf.

So why are you not stripping those couple of files?

Cheers,
Emilio

[2] https://github.com/JuliaLang/julia/issues/23115#issuecomment-320715030