Making fmt private

[sacpis]

Making fmt private.

Fixes #2997.

[bmhowe23]

Thanks for taking a look at this Sachin. Before merging, could you please check what CUDA-QX changes would be needed for this change?

[sacpis]

Thanks for taking a look at this Sachin. Before merging, could you please check what CUDA-QX changes would be needed for this change?

Sure.

[amccaskey]

The big issue here is that Logger.h is a common header for downstream libraries (so they can get cudaq::info()) and it has FmtCore.h as an include. The changes here do not change that. The real fix is to remove any fmtlib headers from our own headers and hide them in cpp implementation code. I'm not sure how one would do that and keep cudaq::info() variadic though.

[boschmitt]

Thanks for taking a look at this Sachin. Before merging, could you please check what CUDA-QX changes would be needed for this change?

CUDA-QX have a couple of files that depends on fmt that comes from CUDA-Q. Thus we would either make those files not depend on fmt or make CUDA-QX have its own private fmt.

[sacpis]

The big issue here is that Logger.h is a common header for downstream libraries (so they can get cudaq::info()) and it has FmtCore.h as an include. The changes here do not change that. The real fix is to remove any fmtlib headers from our own headers and hide them in cpp implementation code. I'm not sure how one would do that and keep cudaq::info() variadic though.

Thanks @amccaskey. I have a question about another option. Why shouldn't we upgrade fmt to the latest version (11.2.0)?

[george-pippos-qb]

Thank you all for taking a look at this for me after I raised this issue. I am trying to plan some work for our company's project that uses CUDA-Q and need to know whether this is likely to be in the next release (and a rough ETA of when the next release will be). I will need to look at alternatives for working around this problem in our repo if timelines don't line up so I would greatly appreciate an update.

[bmhowe23]

Thank you all for taking a look at this for me after I raised this issue. I am trying to plan some work for our company's project that uses CUDA-Q and need to know whether this is likely to be in the next release (and a rough ETA of when the next release will be). I will need to look at alternatives for working around this problem in our repo if timelines don't line up so I would greatly appreciate an update.

It sounds like our primary near-term option is to simply update the fmt library. Do you agree @sacpis?

[sacpis]

Thank you all for taking a look at this for me after I raised this issue. I am trying to plan some work for our company's project that uses CUDA-Q and need to know whether this is likely to be in the next release (and a rough ETA of when the next release will be). I will need to look at alternatives for working around this problem in our repo if timelines don't line up so I would greatly appreciate an update.

It sounds like our primary near-term option is to simply update the fmt library. Do you agree @sacpis?

Yes, I do. Will put up a PR with updated fmt.

[sacpis]

@george-pippos-qb @bmhowe23 Just FYI, we have updated fmt.