* Carlos O'Donell:
> On 4/30/21 12:37 AM, Florian Weimer wrote:
>> * Carlos O'Donell:
>>
>>> On 4/29/21 12:09 PM, Florian Weimer wrote:
>>>> I'd like to hear some feedback on this. It's quite hackish, but
I think
>>>> the complexity is reduced compared to what we had before.
>>>>
>>>> I asked about the macro rewrite on the devel list:
>>>>
>>>>
<
https://lists.fedoraproject.org/archives/list/devel@lists.fedoraproject.o...
>>>>
>>>> My position on the static archives is that if we want debugging
>>>> information for them, that should be a distribution-wide change
>>>> (which I proposed a few years ago, but it was rejected).
>>>>
>>>> Thanks,
>>>> Florian
>>>>
8<------------------------------------------------------------------8<
>>>> This needs a wrapper script around find-debuginfo.sh, so that
>>>> ld.so is not stripped. Overall size for glibc.rpm with and without
>>>> this change is reduced on x86_64:
>>>>
>>>> Before: 16146907
>>>> After: 14396635
>>>>
>>>> This commit resolves debuginfo file conflicts (#1886295) because
>>>> the ldconfig debuginfo file now includes the architecture name.
>>>> All debugging information is preserved in ld.so (#1905611). With
>>>> the distribution defaults, we strip all binaries again (#1661510).
>>>
>>> What size impact does this have on ld.so
>>
>> As I wrote, the proposed change decreases overall installation size.
>> /lib64/ld-2.33.9000.so alone grows from 288360 to 841728 bytes.
>
> OK, so ~550KiB of additional storage for the dynamic loader, but we get
> all the benefits of observability into the loader itself using this
> information. I think the benefit outweighs the costs.
As I said, it's not really additional storage because the install size
is smaller.
> Do you have a scratch build I can look at?
A scratch build is running here:
<
https://koji.fedoraproject.org/koji/taskinfo?taskID=66953022>
I've just started it, so it will take a while to complete.
(We need to backport this all the way to Fedora 33 anyway.)
Thanks,
Florian