[backport] fixes #23711; C code contains backtick`gensym #23716
Conversation
compiler/semtypes.nim
Outdated
| @@ -555,11 +555,15 @@ proc semIdentWithPragma(c: PContext, kind: TSymKind, n: PNode, | |||
| else: discard | |||
| else: | |||
| result = semIdentVis(c, kind, n, allowed) | |||
| let invalidPragmasForPush = if fromTopLevel and "`gensym" notin result.name.s: | |||
There was a problem hiding this comment.
Could we maybe remove the check for `gensym here and mangle the output of {.exportc.} when no string is given to it instead? Because this errors too despite not being skVar/skLet/skConst (#20911):
{.push exportc.}
template foo =
proc bar {.gensym.} = discard
foo()
There was a problem hiding this comment.
Too hacky.
How should we check gensym? There is not a flag dedicated for it since it is a newly created ident.
if sfGenSym in s.flags:
result.add newIdentNode(getIdent(c.ic, x.name.s & "`gensym" & $c.instID),
if c.instLines: actual.info else: templ.info)mangle the output of {.exportc.} when no string is given to it instead
I would like to revive #19636 with some adaptations
There was a problem hiding this comment.
Introduce a new flag like sfWasGenSym.
There was a problem hiding this comment.
Okay, I will move `gensym check into newSymG and incl the flag there.
There was a problem hiding this comment.
At least make the check cheaper by doing:
getIdent(c.ic, "`" & x.name.s & "`gensym" & $c.instID)
...
s.name.s[0] == '`'
There was a problem hiding this comment.
Well, if the name is changed, it is going to break https://github.com/iffy/nim-argparse/blob/4048a76b24919807b0871dd145ba93a48c2688e7/src/argparse/macrohelp.nim#L36 which splits gensym based on "`". Since we are fixing a regression, I will settle for a comprised solution:
if find(result.name.s, '`') >= 0:
result.flags.incl sfWasGenSym
|
|
|
Thanks for your hard work on this PR! Hint: mm: orc; opt: speed; options: -d:release |
This adds Constantine to the important packages. Release announcements: - https://forum.nim-lang.org/t/11935 - https://github.com/mratsim/constantine/releases/tag/v0.1.0 Unfortunately at the moment I'm in a conundrum. - Constantine cannot compile on devel due to #23547 - The workaround is changing ```Nim func mulCheckSparse*(a: var QuadraticExt, b: static QuadraticExt) {.inline.} = ``` to ```Nim template mulCheckSparse*(a: var QuadraticExt, b: QuadraticExt) = ``` but this does not compile on v2.0.8 due to `gensym` issues despite #23716 . i.e. as mentioned in the issue #23711 there is another gensym bug within templates that was fixed in devel but not the v2.0.x series and that is not fixed by #23716
fixes #23711