A quick poll about GHC's -with-rtsopts flag

Who here is: 1) Aware that multiple invocations of -with-rtsopts clobber each other, and 2) Actually relies on this behaviour? This quick and informal poll in relation to this GHC issue: https://gitlab.haskell.org/ghc/ghc/issues/18117 Cheers, Merijn

I’m aware of it because I’ve run into it :p I do not rely on such behavior, though having it be consistently wonky across GHCs might be good... Cheers, Vanessa McHale
On Apr 30, 2020, at 9:21 AM, Merijn Verstraaten
wrote: Who here is: 1) Aware that multiple invocations of -with-rtsopts clobber each other, and 2) Actually relies on this behaviour?
This quick and informal poll in relation to this GHC issue: https://gitlab.haskell.org/ghc/ghc/issues/18117
Cheers, Merijn _______________________________________________ Haskell-Cafe mailing list To (un)subscribe, modify options or view archives go to: http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe Only members subscribed via the mailman list are allowed to post.

On Thu, Apr 30, 2020 at 04:21:47PM +0200, Merijn Verstraaten wrote:
1) Aware that multiple invocations of -with-rtsopts clobber each other, and 2) Actually relies on this behaviour?
Haven't thought about it, but this may makes sense, when configuring options in more than one file, e.g. foo.cabal, and stack.yaml, and whichever precedence, should probably be definitive. Likely cumulative behaviour would be more surprising. [ Based on the assumption that GHC options in stack.yaml are in addition to the component-specific options in the cabal file. ] The name is plural, so takes all the options at once. There could be new options to incrementally modify the list of rtsopts: -set-rtsopt? -add-rtsopt? -del-rtsopt? assuming there's a sensible way to accumulate these. -- Viktor.

The current (i.e. clobbering) behavior seems like the right choice to me,
for what it's worth.
On Thu, Apr 30, 2020 at 11:12 AM Viktor Dukhovni
On Thu, Apr 30, 2020 at 04:21:47PM +0200, Merijn Verstraaten wrote:
1) Aware that multiple invocations of -with-rtsopts clobber each other, and 2) Actually relies on this behaviour?
Haven't thought about it, but this may makes sense, when configuring options in more than one file, e.g. foo.cabal, and stack.yaml, and whichever precedence, should probably be definitive. Likely cumulative behaviour would be more surprising.
[ Based on the assumption that GHC options in stack.yaml are in addition to the component-specific options in the cabal file. ]
The name is plural, so takes all the options at once. There could be new options to incrementally modify the list of rtsopts:
-set-rtsopt? -add-rtsopt? -del-rtsopt?
assuming there's a sensible way to accumulate these.
-- Viktor. _______________________________________________ Haskell-Cafe mailing list To (un)subscribe, modify options or view archives go to: http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe Only members subscribed via the mailman list are allowed to post.
participants (4)
-
Merijn Verstraaten
-
Travis Whitaker
-
Vanessa McHale
-
Viktor Dukhovni