Build size regressions

[pr-benchmarks]

A publish size regression has been detected for the following scenarios:

Scenario	Environment	Date	Old Build (KB)	New Build (KB)	Change	Deviation	StDev
json	Json CG2 composite aspnet-citrine-amd	02/19/2021 08:34:23	151,273	328,772	117 % (177,499) 👎	12,083.99 σ	15
json	Json CG2 non-composite aspnet-citrine-amd	02/19/2021 08:35:31	123,245	300,724	144 % (177,479) 👎	15,876.74 σ	11
json	Json CG2 composite avx2 aspnet-citrine-amd	02/19/2021 08:36:59	151,786	329,285	116 % (177,499) 👎	11,823.81 σ	15

Changes:

Runtime	Previous Version	Current Version	Diff
ASP.NET	n/a	n/a	n/a
CLR	n/a	n/a	n/a
SDK	6.0.100-preview.2.21120.3	6.0.100-preview.3.21123.18	n/a

@sebastienros
@mangod9

[sebastienros]

Regression in build size impact all self-contained apps, isolated with

3e85099...1db079d

| .NET Runtime Version  | 6.0.0-preview.2.21118.6+3e85099234a06ff36915b0c11289bcfe6787fc66 |
| .NET Runtime Version  | 6.0.0-preview.2.21118.10+1db079d19538a7520ec6c74f0566b4b78c7e935d |

[sebastienros]

Before

After

[sebastienros]

Commands to repro:

crank --config https://raw.githubusercontent.com/aspnet/Benchmarks/master/scenarios/platform.benchmarks.yml --scenario json --profile aspnet-perf-lin --variable warmup=0 --variable duration=1 --application.framework net6.0 --application.runtimeversion 6.0.0-preview.2.21118.6

crank --config https://raw.githubusercontent.com/aspnet/Benchmarks/master/scenarios/platform.benchmarks.yml --scenario json --profile aspnet-perf-lin --variable warmup=0 --variable duration=1 --application.framework net6.0 --application.runtimeversion 6.0.0-preview.2.21118.10

Add --application.options.fetch true to download the published apps

[mangod9]

moving to VM for now, looks like all native binaries increased in size.

[mangod9]

FYI @jkoritzinsky @trylek @jkotas if they are aware of any build changes around symbols not getting stripped or something similar?

[EgorBo]

Can it be #48444 from that list of commits?
cc @am11

[jkotas]

Can it be #48444 from that list of commits?

I agree that this is the problematic change. cc @janvorli

I think we should revert it. Thoughts? Also, we may want to push the revert into preview2 since the bug is there as well.

[jkotas]

#49092

[jkotas]

Fixed by #49092