schannel: reclassify extra-verbose schannel_recv messages

[jay]

• Create a new macro SCH_DEV() to manage verbose debug messages that are only useful for debugging Schannel recv decryption.

schannel_recv contains a lot of useful debug messages to help debug the function, however in practice they are not otherwise useful and showing them in debug builds adds a lot of noise.

To show these messages curl must now be built with CURL_SCHANNEL_DEV_DEBUG defined.

Prior to this change many, but not all, extra-verbose messages were wrapped in DEBUGF() so they were only shown in debug builds.

Ref: #14807

Closes #xxxxx

[dfandrich]

Analysis of PR #14826 at 03d9f837:

Test 1540 failed, which has NOT been flaky recently, so there could be a real issue in the PR.

Test 677 failed, which has NOT been flaky recently, so there could be a real issue in the PR.

Generated by Testclutch

[jay]

Test 1540 failed, which has NOT been flaky recently, so there could be a real issue in the PR.

This test failed in the Omni OS ci job and I don't see anything unexpected in the output except the return code is strange lib1540 returned 2006, when expecting 0. As far as I can see that can't be returned (see lib1540.c + first.c) unless curl_easy_perform returns it so I think more likely there must be something strange happening in runtests

edit: oops I missed in the output that there was a crash so that explains it

2024-09-08T07:00:17.9651993Z  Assertion failed: 0, file ../../lib/multi.c, line 1131, function multi_getsock
2024-09-08T07:00:17.9652853Z  sh: 5582: Abort(coredump)

Test 677 failed, which has NOT been flaky recently, so there could be a real issue in the PR.

another unknown error code lib677.exe returned 143, when expecting 0 despite expected output, this one in CI job msvc, CM x64-windows libressl.

edit: looked at this one again and don't see anything wrong in the output

[jay]

@vszakats do you have advice for how I should append a definition to cmake CFLAGS for a CI job, I tried -DCMAKE_C_FLAGS_INIT:STRING=-DCURL_SCHANNEL_DEV_DEBUG but CMAKE_C_FLAGS_INIT was overridden by a CMAKE_C_FLAGS which is used for all jobs.

https://github.com/curl/curl/blob/12b269fe597992703ac3c0c1284ca129ccbf9366/.github/workflows/windows.yml#L176

So then I tried -DCMAKE_C_FLAGS:STRING=-DCURL_SCHANNEL_DEV_DEBUG but it seems cmake doesn't work that way (it only uses the last CMAKE_C_FLAGS as far as I can tell), which means it will add -DCURL_SCHANNEL_DEV_DEBUG overriding the CMAKE_C_FLAGS used for all jobs (shown above with -DTESTING_CI_BEHAVIOR which is what is overridden in this experiment)

https://github.com/curl/curl/blob/12b269fe597992703ac3c0c1284ca129ccbf9366/.github/workflows/windows.yml#L62

[vszakats]

@vszakats do you have advice for how I should append a definition to cmake CFLAGS for a CI job, I tried -DCMAKE_C_FLAGS_INIT:STRING=-DCURL_SCHANNEL_DEV_DEBUG but CMAKE_C_FLAGS_INIT was overridden by a CMAKE_C_FLAGS which is used for all jobs.

https://github.com/curl/curl/blob/12b269fe597992703ac3c0c1284ca129ccbf9366/.github/workflows/windows.yml#L176

So then I tried -DCMAKE_C_FLAGS:STRING=-DCURL_SCHANNEL_DEV_DEBUG but it seems cmake doesn't work that way (it only uses the last CMAKE_C_FLAGS as far as I can tell), which means it will add -DCURL_SCHANNEL_DEV_DEBUG overriding the CMAKE_C_FLAGS used for all jobs (shown above with -DTESTING_CI_BEHAVIOR which is what is overridden in this experiment)

https://github.com/curl/curl/blob/12b269fe597992703ac3c0c1284ca129ccbf9366/.github/workflows/windows.yml#L62

Yes, only the last one counts. Would this help?:

diff --git a/.github/workflows/windows.yml b/.github/workflows/windows.yml
index 9ce3c7650..500e671f2 100644
--- a/.github/workflows/windows.yml
+++ b/.github/workflows/windows.yml
@@ -171,6 +171,7 @@ jobs:
           fi
           [ '${{ matrix.type }}' = 'Debug' ] && options+=' -DCMAKE_RUNTIME_OUTPUT_DIRECTORY_DEBUG='
           [ '${{ matrix.type }}' = 'Release' ] && options+=' -DCMAKE_RUNTIME_OUTPUT_DIRECTORY_RELEASE='
+          cflags+=' -DCURL_SCHANNEL_DEV_DEBUG'
           cmake -B bld ${options} \
             "-DCMAKE_C_FLAGS=${cflags}" \
             "-DCMAKE_RC_COMPILE_OBJECT=${rcopts}" \
@@ -389,6 +390,7 @@ jobs:
         if: ${{ matrix.build == 'cmake' }}
         run: |
           cmake -B bld \
+            -DCMAKE_C_FLAGS=-DCURL_SCHANNEL_DEV_DEBUG \
             -DCMAKE_SYSTEM_NAME=Windows \
             -DCMAKE_C_COMPILER_TARGET=${TRIPLET} \
             -DCMAKE_C_COMPILER=${TRIPLET}-gcc \
@@ -502,6 +504,7 @@ jobs:
             options+=" -DOPENSSL_ROOT_DIR=$VCPKG_INSTALLATION_ROOT/installed/${{ matrix.arch }}-${{ matrix.plat }}"
           fi
           cmake -B bld ${options} \
+            -DCMAKE_C_FLAGS=-DCURL_SCHANNEL_DEV_DEBUG \
             "-DCMAKE_TOOLCHAIN_FILE=$VCPKG_INSTALLATION_ROOT/scripts/buildsystems/vcpkg.cmake" \
             "-DVCPKG_INSTALLED_DIR=$VCPKG_INSTALLATION_ROOT/installed" \
             '-DVCPKG_TARGET_TRIPLET=${{ matrix.arch }}-${{ matrix.plat }}' \

[jay]

@vszakats i want the define only for job 'schannel dev debug' so i changed the 'cmake configure' step to add the cflags from that particular job like this

-            "-DCMAKE_C_FLAGS=${cflags}" \
+            "-DCMAKE_C_FLAGS=${{ matrix.cflags }} ${cflags}" \

seems to work ok let me know if you see any problem with it

[vszakats]

It looks fine to me.

Maybe moving the new job one line up, and moving the ENABLE_UNICODE option before the added cmake option would make the new job line up better with existing ones, but it's just a formatting nit.

[bagder]

@jay where are we on this?

[jay]

Some Windows CI failures I can't explain and probably unrelated. I rebased on master and force pushed. Assuming @icing is OK with it then it can replace #14810 and I'll land it

[jay]

@icing

[icing]

Sorry, missed the mention during covid. Looks good to me, thanks!