fix #13412 nim now recompiles for stdin input; SuccessX now configurable; can show whether it recompiled

[timotheecour]

• fix echo 'echo 1' | nim c -r - silently gives wrong results (nimBetterRun not updated for stdin) #13412
  (in future PR this could be improved by not recompilining if stdin was the same but not sure how to do that; help welcome)
• SuccessX now configurable; can show whether it recompiled

note

IMO we should simply not support reading from stdin via pipes, Nim is a compiler, it's not for throwaway code.

it's very useful and I use it a lot, lots of modern languages support it, eg python (python -c), D (rdmd --eval) and plenty others
this is very useful when you just want to quickly test a single command without having to open a file just for that;
having 1 liners that can be copy pasted in a shell without further work is quite useful too

Besides, there is already nim secret for that.

that's a different beast, VM code semantics is different from runtime code and there are other differences

[Araq]

The patch to extccomp is ok, the rest is noisy. How about we report 0 for LOC if it's a cached build? It makes sense, right, if it's cached, we looked at 0 lines of code.

[kaushalmodi]

I stumbled across this today..

I meant to highlight above that echo 'hello' | .. would have failed the nim compilation (because it should have been echo 'echo "hello"' | ..). But it still printed the output of the last successfully compiled stdinfile binary.

Even with corrected echo 'echo "hello"' | .., it still printed the old output. i.e. the stdinfile doesn't get recompiled.

[timotheecour]

How about we report 0 for LOC if it's a cached build? It makes sense, right, if it's cached, we looked at 0 lines of code.

nah, it's not 0 because of various config files etc; that would be a lie and also not what user wants

The patch to extccomp is ok, the rest is noisy.

found a much better future proof solution, now I can add this to my ~/.config/nim/config.nims (or just use cmd line but that's not practical for that):

--msgs.hintSuccessX:customMsg
eg:
--msgs.hintSuccessX:"$loc LOC; $sec sec; $mem; $build build; proj: $project; out: $output recompiled: $projectRecompiled"

and it works exactly the way user wants, without affecting others, and can be later augmented with other fields a specific user may care about
I'm leaving it undocumented for now.

note:

using --msgs.hintSuccessX instead of --msgs.successX since it makes it easier to search in compiler sources

[timotheecour]

ping @Araq (just rebased, no changes, was green already before rebase)

[Clyybber]

@timotheecour

The patch to extccomp is ok, the rest is noisy.

Can you restrict this PR to the extccomp patch?

[timotheecour]

Can you restrict this PR to the extccomp patch?

done, PTAL (I will revisit the other changes I was making here in another PR as it is useful, there are still existing bugs related to nim not rebuilding when it should, which forces the suboptimal -f in some situations)