Regarding the "More Stuff" comment: Yes. Of course. But "small steps" and "increments" and "Rome != oneDay".

PS To a first approximation each single bullet up there is a PR and reverse depends check. Some packages still useRcppLdFlags() which was made redundant maybe a dozen years ago. You have to careful when removing.

We cannot address all possible changes in a PR. I like the amount of change and hence potential instability here and will likely try to draw a dotted and test this at the current stage via a reverse depends run. And then proceed with, say, a variadic template if/else cleanup at a time hopefully leading to one coherent PR. Does that work for you?

We cannot address all possible changes in a PR.

Sure, just documenting further steps that I've seen while looking around, which, as I said, can be part of other PRs.

Maybe we should think about using the wiki, or discussions, or .... $INSERT_WHATEVER_HERE ... as comments deeply nested in one of hundreds of PR tend to get lost.

If#1363 is supposed to cover several PRs, I can move the comments there.

That may work. I usually think of a 'one issue -- one pr' correspondance but we don't have to.

Or maybe even open anew issue just to hold comments for possibly multiple issues and prs? Cleaner still?

Sad to report that I am seeing a lot of 'carnage'. About eight hours in (after I did wait for the previous, more innocent PR to finish a reverse depends as 1.0.14.6) I see 373 successes but 14 failures (and the usual hardwired 4 skipped at this point). Failures are

Failed packages:  agcounts, alcyon, aRtsy, asteRisk, ASV, BAMBI, bamm, bayesianVARs, BayesMallows, BGVAR,BHSBVAR, bioacoustics, cbbinom, CERFI

and we have errors such as a simple brace initialization failing which I just used to bissect:

Rscript -e'Rcpp::cppFunction("Rcpp::NumericVector makevec(int n) { Rcpp::NumericVector z = {1,2,3}; return z; }")'

With the ten commits here being

12c29eee * origin/feature/header_cleanup Following rebase roll micro release and date, update ChangeLoge1150dc5 * Reflect code review comments8809c39b * Remove HAS_STATIC_ASSERT define and simplify at its sole use714c4834 * Additional simplification, and typo fixd4870afd * Reduced compiler.h to about a page of mostly defines and includesd2c859aa * Further cleanup39f3c422 * No longer provide fallback for unordered_{set,map} which are given2665b4d3 * Expand macos matrix to macos-13, simplify via r-ci with bootstrapf1956608 * Do not rely on __has_feature() for g++1d9a5497 * First step of simplifying compiler checks

I can pinpoint the break betweend4870af (still good) and714c483 (no longer compiles simple example). I may stop the reverse depends and re-start there.

So this incomplete rev dep run was under the (hoped for) 1.0.14.7 version; the last good one was 1.0.14.6 after the last PR. I am setting up a new one at the pinpointed bisection as 1.0.14.6.1.

Incidentally I can cherry-pick the remaining PRs back, omitting just one commit from the sequence, and that passes the local test. So this may be a valid candidate too, tentatively named 1.0.14.6.2. I am working off a 'repair' branch called feature/header_cleanup_reset which I may push. It now has

edd@rob:~/git/rcpp(feature/header_cleanup_reset)$ git ls| head -10* 7f309e17 - (HEAD -> feature/header_cleanup_reset) Following rebase roll micro release and date, update ChangeLog (6 minutes ago)<Dirk Eddelbuettel>* de01352d - Reflect code review comments (7 minutes ago)<Dirk Eddelbuettel>* e87c738f - Remove HAS_STATIC_ASSERT define and simplify at its sole use (10 minutes ago)<Dirk Eddelbuettel>* d4870afd - Reduced compiler.h to about a page of mostly defines and includes (9 hours ago)<Dirk Eddelbuettel>* d2c859aa - Further cleanup (9 hours ago)<Dirk Eddelbuettel>* 39f3c422 - No longer provide fallbackfor unordered_{set,map} which are given (9 hours ago)<Dirk Eddelbuettel>* 2665b4d3 - Expand macos matrix to macos-13, simplify via r-ci with bootstrap (9 hours ago)<Dirk Eddelbuettel>* f1956608 - Do not rely on__has_feature()for g++ (9 hours ago)<Dirk Eddelbuettel>* 1d9a5497 - First step of simplifying compiler checks (9 hours ago)<Dirk Eddelbuettel>* 0905d92e - (origin/master, origin/HEAD, master) Use lsInteral3 instead of lsInternal (#1362) (10 hours ago) <Dirk Eddelbuettel>edd@rob:~/git/rcpp(feature/header_cleanup_reset)$

which is the above up tod4870af, omitting the identified714c483 and then re-adding the remaining three (via cherry-pick so yields new sha1).

Oh I am smelling the bacon -- I overlooked one use of one of those three defines:

That would do it. Seems like I can resurrect this branch and PR.

PS Now 2 1/2 hours in with 120+ packages passed and on issues so far.

PPS Now 12 hours in, 600+ packages, zero issues so far. Looks good.

This can likely be merged in a day or two once the run is done to form the anticipated 'base station' from which we can launch further targeted cleanups, notably simplifying the use of variatic templates. I should be able to get to that by mid-week.