Runtime test output transcripts in the xml result files use excessive escaping rendering them basically unreadable #92092
Description
The result xml file generated by the new-style merged test wrappers includes stdout / stderr transcript for each test. I believe @jkoritzinsky initially implemented this functionality with the aim to help developers directly see the output transcript from the tests as an aid in issue investigation. At some point we modified this logic so that the output transcript is so heavily escaped it's basically unreadable, e.g. here:
<test name="readytorun\determinism\crossgen2determinism\crossgen2determinism.cmd" type="OutOfProcessTest" method="readytorun\determinism\crossgen2determinism\crossgen2determinism.cmd" time="11.268134" result="Pass">
<output>
<![CDATA[ _x000D__x000A_Return_x0020_code:_x0020__x0020__x0020__x0020__x0020__x0020_0_x000D__x000A_Raw_x0020_output_x0020_file:_x0020__x0020__x0020__x0020__x0020__x0020_D:_x005C_git_x005C_runtime3_x005C_artifacts_x005C_tests_x005C_coreclr_x005C_windows.x64.Checked_x005C_readytorun_x005C_determinism_x005C_crossgen2determinism_x005C_output.txt_x000D__x000A_Raw_x0020_output:_x000D__x000A_BEGIN_x0020_EXECUTION_x000D__x000A__x000D__x000A_D:_x005C_git_x005C_runtime3_x005C_artifacts_x005C_tests_x005C_coreclr_x005C_windows.x64.Checked_x005C_readytorun_x005C_determinism_x005C_crossgen2determinism_x003E_set_x0020_DOTNETCLI_x003D_D:_x005C_git_x005C_runtime3_x005C_dotnet.cmd_x0020__x000D__x000A_Emitting_x0020_R2R_x0020_PE_x0020_file:_x0020_crossgen2smoke1.ildll_x000D__x000A_Emitting_x0020_map_x0020_file:_x0020_crossgen2smoke1.map_x000D__x000A_Emitting_x0020_R2R_x0020_PE_x0020_file:_x0020_crossgen2smoke2.ildll_x000D__x000A_Emitting_x0020_map_x0020_file:x0020_crossgen2smoke2.map_x000D__x000A__x0020__x0022_D:x005C_git_x005C_runtime3_x005C_artifacts_x005C_tests_x005C_coreclr_x005C_windows.x64.Checked_x005C_Tests_x005C_Core_Root_x005C_corerun.exe_x0022__x0020-p_x0020__x0022_System.Reflection.Metadata.MetadataUpdater.IsSupported_x003D_false_x0022__x0020-p_x0020__x0022_System.Runtime.Serialization.EnableUnsafeBinaryFormatterSerialization_x003D_true_x0022__x0020__x0020_crossgen2determinism.dll_x0020__x000D__x000A_Files_x0020_of_x0020_length_x0020_139264_x0020_were_x0020_identical._x000D__x000A_Expected:_x0020_100_x000D__x000A_Actual:x0020_100_x000D__x000A_END_x0020_EXECUTION_x0020-x0020_PASSED_x000D__x000A_PASSED_x000D__x000A_Test_x0020_Harness_x0020_Exitcode_x0020_is_x0020:_x0020_0_x000D__x000A_To_x0020_run_x0020_the_x0020_test:_x000D__x000A_Set_x0020_up_x0020_CORE_ROOT_x0020_and_x0020_run._x000D__x000A__x003E__x0020_D:x005C_git_x005C_runtime3_x005C_artifacts_x005C_tests_x005C_coreclr_x005C_windows.x64.Checked_x005C_readytorun_x005C_readytorun_x005C..x005C_determinism_x005C_crossgen2determinism_x005C_crossgen2determinism.cmd_x000D__x000A ]]>
</output>
</test>
We seem to be escaping totally benign characters including spaces, CR/LF pairs or backslashes, increasing the size of the result files and reducing their value for developers, frankly speaking it would probably take me half an hour or so to turn the above into something reasonable. Jeremy speculated that we may have broken this with the xml checking / completion that was part of the Merge on Red project, assigning to Ivan to take a first look and we'll follow from there.
Thanks
Tomas
/cc @dotnet/runtime-infrastructure
P.S. I have spent several iterations trying to figure out how to format the above thing in some reasonable manner. I failed miserably.