Prepare release v.1.9.0 (tentative: 3rd March)

[lalitb]

There are few Open PRs which preferably be merged (not in order of priority) before creating the release:

• [EXPORTER] Delegate all API calls of gRPC into opentelemetry_exporter_otlp_grpc_client, and make it contains all symbols needed. #2005
• Add attributes for InstrumentationScope #2004
• [SDK] Add ForceFlush for all LogRecordExporters and SpanExporters. #2000
• [Metrics SDK] Performance improvement in measurement processing #1993
• Opentracing shim #1909

#2000 is ABI breaking change for the SDK which should be fine in general. But it also means that if someone is using otel-cpp-sdk*so targets with their application, the application can crash if these shared library is rebuild for new version, without recompiling/relinking the application against these new shared libraries. This needs to be mentioned in the release notes.

[marcalff]

Please revert the fix done with PR: Rename the global SDK version variables to avoid naming clash #2011 .

OPENTELEMETRY_SDK_MAJOR_VERSION needs to be a pre processor symbol to fix issue Provide major/minor/patch version macros #2012 .

[marcalff]

@lalitb @ThomsonTan @esigo @owent Please see above comment.

[ThomsonTan]

Please revert the fix done with PR: Rename the global SDK version variables to avoid naming clash #2011 .

OPENTELEMETRY_SDK_MAJOR_VERSION needs to be a pre processor symbol to fix issue Provide major/minor/patch version macros #2012 .

Is the revert required? I think the PR #2011 is separate the the ask of #2012 which looks like a feature request? The PR was targeting a different issue of name clashing.

[owent]

I found some commits are missed in #2005,Can I push them next monday?I added a CI job and a unit test to check #1603.

[ThomsonTan]

I found some commits are missed in #2005,Can I push them next Monday added a CI job and a unit test to check #1603.

It looks good to me. We can wait it before completing the release. @lalitb @marcalff @esigo how do you think?

[marcalff]

Please revert the fix done with PR: Rename the global SDK version variables to avoid naming clash #2011 .
OPENTELEMETRY_SDK_MAJOR_VERSION needs to be a pre processor symbol to fix issue Provide major/minor/patch version macros #2012 .

Is the revert required? I think the PR #2011 is separate the the ask of #2012 which looks like a feature request? The PR was targeting a different issue of name clashing.

I think the revert is required.

Code introduced:

namespace sdk
{
namespace version
{
extern const int OPENTELEMETRY_SDK_MAJOR_VERSION;

will break once pre processor symbol is added for #2012 :

#define OPENTELEMETRY_SDK_MAJOR_VERSION 1

[ThomsonTan]

OPENTELEMETRY_SDK_MAJOR_VERSION

Ok, but since as the preprocessor symbol #define OPENTELEMETRY_SDK_MAJOR_VERSION is new, probably we can name it as OPENTELEMETRY_CPP_SDK_MAJOR_VERSION? We cannot use the global var name MAJOR_VERSION anyway as it is causing more general clashing.

[lalitb]

The intended purpose of the existing const variables was to stamp the version information in the binary.

$strings /usr/local/lib/libopentelemetry_version.a  | grep 1.8.2
pre_release_1.8.2
1.8.2
1.8.2-NONE-NONE-37-pre_release_1.8.2-435ce60f233b6718aaa04bb7068dd641b536299b

We shouldn't be removing them, or replacing them with macros. Macros should be introduced separately for #2012. Though it should be fine to change these variable names to lowercase (or something else) in case of conflict.

[marcalff]

I agree these const variables should be available in the library, this is not the problem.

The problem is that resolving one name clash for MAJOR_VERSION created another name clash instead.

Ok to change to lowercase instead:

namespace sdk
{
namespace version
{
extern const int major_version;

This avoid the conflict with MAJOR_VERSION, and with OPENTELEMETRY_SDK_MAJOR_VERSION anticipated for #2012

[marcalff]

@ThomsonTan @lalitb

So, if sdk::version::major_version looks ok to everyone, no need for a full revert, but some adjustments are needed instead before publishing the release.

[ThomsonTan]

Ok. I'll make a separate PR to change the C++ version variables into lower case.