Hotfix: fixes CI by correcting cpp-logger and DLIO profiler/dftracer stuff #117
Conversation
ilumsden
force-pushed
the
cpp_logger_find_package_fix
branch
from
dc75990 to
c2a17a6
Compare
ilumsden
changed the title
|
@ilumsden We will have to switch DYAD_PROFILER macros to use DFTracer as well. |
|
Good point. I was only thinking about it from the Python side since that was the only place where things were failing in CI. But, if I'm going to make these changes in the Python API, I should do it in the main body of DYAD too. I'll work on that after lunch |
|
I recommend doing a search on DLIO_PROFILER and replacing it with DFTRACER. You will have macros, apis, env variables. |
|
@hariharan-devarajan I made all the changes you requested. |
src/dyad/core/dyad_ctx.c
Outdated
| @@ -286,24 +289,25 @@ dyad_rc_t dyad_init (bool debug, | |||
| } | |||
|
|
|||
| ctx->relative_to_managed_path = relative_to_managed_path; | |||
| DYAD_LOG_INFO (ctx, "DYAD_CORE INIT: relative_to_managed_path %s", \ | |||
| DYAD_LOG_INFO (ctx, | |||
There was a problem hiding this comment.
This will end up in a compilation error.
There was a problem hiding this comment.
I updated the formatting changes
src/dyad/modules/dyad.c
Outdated
| "DYAD_MOD: Failed to load file \"%s\" only read %zd of %zd.", \ | ||
| fullpath, \ | ||
| inlen, \ | ||
| DYAD_LOG_ERROR (mod_ctx->ctx, |
There was a problem hiding this comment.
I updated the formatting changes
src/dyad/modules/dyad.c
Outdated
| if (flux_respond_error (h, msg, ENODATA, NULL) < 0) { | ||
| DYAD_LOG_ERROR (mod_ctx->ctx, "DYAD_MOD: %s: flux_respond_error with ENODATA failed\n", __func__ ); | ||
| DYAD_LOG_ERROR (mod_ctx->ctx, |
There was a problem hiding this comment.
I updated the formatting changes
|
There are many needless styling modifications in this PR. Refrain from that in the future and focus on what the PR intends to fix. |
While looking at CI for other PRs, I found that GitHub Actions was failing to build DYAD because it couldn't find cpp-logger. After looking into the issue, I realized that our check for the
<PackageName>_FOUNDvariable afterfind_packagewas simply using the wrong variable name. It was usingCPP_LOGGER_FOUND, while it should have been usingcpp-logger_FOUND. This PR corrects that variable name.After fixing this issue with cpp-logger, I also discovered that CI was failing due to @hariharan-devarajan's recent renaming of DLIO profiler to dftracer. This PR also modifies the Python bindings to use the new name (and other minor API changes in dftracer).