On Fri, Oct 21, 2016 at 02:23:47PM +0100, Peter Robinson wrote:
On Fri, Oct 21, 2016 at 2:21 PM, Zbigniew Jędrzejewski-Szmek
<zbyszek(a)in.waw.pl> wrote:
> On Fri, Oct 21, 2016 at 01:31:25PM +0200, Jan Kratochvil wrote:
>> On Fri, 21 Oct 2016 13:18:38 +0200, Peter Robinson wrote:
>> > guild would be because it's a dep of a dep of gdb-headless
>> guile
>>
>> libguile-2.0.so.22 is DT_NEEDED - as shown by ldd.
>>
>> Easy way would be to make gdb-headless a separate binary/build.
>>
>> Less easy way would be to dlopen() libguile from gdb and keep there some stub
>> with dlsym()ed pointers to functions. Or maybe provide weak symbols all
>> pointing to a function dlopen()ing libguile and so the weak symbols would get
>> overriden by real symbols from libguile. Or is solved by some project?
>>
>> Not sure if that guile dependency is such an issue.
>
> Can't we instead add fake Provides: this-package-is-not-critpath
> and ignore such packages from the script which makes them critpath?
> This seems like a better solution than doing ugly things like dlopen
> (and breaking automatic Requires, etc.)
Or just not care if they're critpath? I'm not sure what the problem is.
Additional constraints on updates.
Zbyszek