ResponseCompressionMiddlewareTests failing in release/8.0 on Fedora41

[wtgodbe]

Failing Test(s)

• Failed Microsoft.AspNetCore.ResponseCompression.Tests.ResponseCompressionMiddlewareTest.SendFileAsync_AfterFirstWrite_CompressesAndFlushes

Error Message

2025-01-14T10:02:12.7516868Z Expected: 46
2025-01-14T10:02:12.7516920Z Actual:   43

Stacktrace

2025-01-14T10:02:12.7532953Z      at Microsoft.AspNetCore.ResponseCompression.Tests.ResponseCompressionMiddlewareTest.CheckResponseCompressed(HttpResponseMessage response, Nullable`1 expectedBodyLength, String expectedEncoding) in /_/src/Middleware/ResponseCompression/test/ResponseCompressionMiddlewareTest.cs:line 1303
2025-01-14T10:02:12.7533052Z    at Microsoft.AspNetCore.ResponseCompression.Tests.ResponseCompressionMiddlewareTest.SendFileAsync_AfterFirstWrite_CompressesAndFlushes() in /_/src/Middleware/ResponseCompression/test/ResponseCompressionMiddlewareTest.cs:line 1167

Logs

Build

https://dev.azure.com/dnceng-public/public/_build/results?buildId=916632&view=results

Other tests such as Failed Microsoft.AspNetCore.ResponseCompression.Tests.ResponseCompressionMiddlewareTest.SendFileAsync_FirstWrite_CompressesAndFlushes fail in a similar way:

2025-01-14T10:02:13.9643316Z Expected: 34
2025-01-14T10:02:13.9643333Z Actual: 31

Likely related to the recent update from Fedore38 to Fedora41

[wtgodbe]

CC @adityamandaleeka new 8.0 test failures on Fedora41

[BrennanConroy]

Compression (at least zlib probably other algorithms too) don't have a guaranteed size between algorithm versions. We either do the simple "fix" of changing the expected size, or we try to be a little looser in our checks by checking a range potentially.

e.g.
Assert.InRange(compressedSize, 20, 40);

[wtgodbe]

Did we already make that change in main? I don't see the same failures there

[BrennanConroy]

I remember we did the lazy fix during a runtime update when the runtime changed the zlib library. It might not be affected by the OS change in the same way.

[wtgodbe]

Looks like that was #56684 - that doesn't touch SendFileAsync_AfterFirstWrite_CompressesAndFlushes though, so it's not obvious to me why we're not seeing failure on main

[BrennanConroy]

runtime changed the zlib library. It might not be affected by the OS change in the same way.

[wtgodbe]

Should we just update the 8.0 tests to use a range then? Are you the right person to make the fix?