Friends Is it my imagination or has Hadrian become a LOT slower? I am trying to run a single test, thus: hadrian/build -j6 \ --flavour=perf+no_profiled_libs+ticky_ghc \ test \ --skip=_build/stage0/** \ --skip=_build/stage1/** \ --skip=_build/stage1/rts/** \ --skip=_build/stage1/libraries/** \ --skip=_build/stage1/utils/** \ --docs=none \ --only=ClassOperator This takes 70 seconds, of which almost all is spent with Hadrian saying stuff below *Something must be wrong! * Seventy seconds to run one test! Then to accept the new output is another 70 seconds. If I have a couple of dozen test-output wibbles to fix, that adds up to a lot of time. Simon ]0;Starting...(/home/simonpj/code/HEAD-31) ]0;Running for 5s [119/38037], predicted 264m08s (0%)(/home/simonpj/code/HEAD-31) ]0;Running for 10s [8853/38038], predicted 36m59s (0%)(/home/simonpj/code/HEAD-31) ]0;Running for 15s [12109/38038], predicted 9m46s (2%)(/home/simonpj/code/HEAD-31) ]0;Running for 20s [13119/38038], predicted 7m03s (4%)(/home/simonpj/code/HEAD-31) ]0;Running for 25s [13879/38038], predicted 7m19s (5%)(/home/simonpj/code/HEAD-31) ]0;Running for 30s [14485/38038], predicted 6m58s (6%)(/home/simonpj/code/HEAD-31) ]0;Running for 35s [15338/38038], predicted 6m06s (8%)(/home/simonpj/code/HEAD-31) ]0;Running for 40s [16434/38038], predicted 2m56s (17%)(/home/simonpj/code/HEAD-31) ]0;Running for 45s [20588/38038], predicted 3m32s (16%)(/home/simonpj/code/HEAD-31) ]0;Running for 50s [26354/38038], predicted 63m21s (1%)(/home/simonpj/code/HEAD-31) ]0;Running for 55s [30040/38038], predicted 49m36s (1%)(/home/simonpj/code/HEAD-31) ]0;Running for 1m00s [35123/38038], predicted 19m59s (4%)(/home/simonpj/code/HEAD-31) # ghc
TBH that doesn't sound like hadrian is slow, it sounds like the test harness or the tests themselves (maybe RTS startup/shutdown?) are and hadrian's giving you the play-by-play from it. On Sun, Sep 27, 2026 at 4:37 PM Simon Peyton Jones via ghc-devs < ghc-devs@haskell.org> wrote:
Friends
Is it my imagination or has Hadrian become a LOT slower?
I am trying to run a single test, thus:
hadrian/build -j6 \ --flavour=perf+no_profiled_libs+ticky_ghc \ test \ --skip=_build/stage0/** \ --skip=_build/stage1/** \ --skip=_build/stage1/rts/** \ --skip=_build/stage1/libraries/** \ --skip=_build/stage1/utils/** \ --docs=none \ --only=ClassOperator
This takes 70 seconds, of which almost all is spent with Hadrian saying stuff below
*Something must be wrong! *
Seventy seconds to run one test! Then to accept the new output is another 70 seconds. If I have a couple of dozen test-output wibbles to fix, that adds up to a lot of time.
Simon
]0;Starting...(/home/simonpj/code/HEAD-31) ]0;Running for 5s [119/38037], predicted 264m08s (0%)(/home/simonpj/code/HEAD-31) ]0;Running for 10s [8853/38038], predicted 36m59s (0%)(/home/simonpj/code/HEAD-31) ]0;Running for 15s [12109/38038], predicted 9m46s (2%)(/home/simonpj/code/HEAD-31) ]0;Running for 20s [13119/38038], predicted 7m03s (4%)(/home/simonpj/code/HEAD-31) ]0;Running for 25s [13879/38038], predicted 7m19s (5%)(/home/simonpj/code/HEAD-31) ]0;Running for 30s [14485/38038], predicted 6m58s (6%)(/home/simonpj/code/HEAD-31) ]0;Running for 35s [15338/38038], predicted 6m06s (8%)(/home/simonpj/code/HEAD-31) ]0;Running for 40s [16434/38038], predicted 2m56s (17%)(/home/simonpj/code/HEAD-31) ]0;Running for 45s [20588/38038], predicted 3m32s (16%)(/home/simonpj/code/HEAD-31) ]0;Running for 50s [26354/38038], predicted 63m21s (1%)(/home/simonpj/code/HEAD-31) ]0;Running for 55s [30040/38038], predicted 49m36s (1%)(/home/simonpj/code/HEAD-31) ]0;Running for 1m00s [35123/38038], predicted 19m59s (4%)(/home/simonpj/code/HEAD-31) # ghc _______________________________________________ ghc-devs mailing list -- ghc-devs@haskell.org To unsubscribe send an email to ghc-devs-leave@haskell.org
-- brandon s allbery kf8nh allbery.b@gmail.com
I can't reproduce this on master. It seems to run in ~17s there which is not great but roughly in line with what I would expect. I agree with Brandon that it seems more likely something in your tree (test, ghc, something else) introduced the slowdown. On 27/09/2026 23:07, Brandon Allbery via ghc-devs wrote:
TBH that doesn't sound like hadrian is slow, it sounds like the test harness or the tests themselves (maybe RTS startup/shutdown?) are and hadrian's giving you the play-by-play from it.
On Sun, Sep 27, 2026 at 4:37 PM Simon Peyton Jones via ghc-devs <ghc-devs@haskell.org> wrote:
Friends
Is it my imagination or has Hadrian become a LOT slower?
I am trying to run a single test, thus:
hadrian/build -j6 \ --flavour=perf+no_profiled_libs+ticky_ghc \ test \ --skip=_build/stage0/** \ --skip=_build/stage1/** \ --skip=_build/stage1/rts/** \ --skip=_build/stage1/libraries/** \ --skip=_build/stage1/utils/** \ --docs=none \ --only=ClassOperator
This takes 70 seconds, of which almost all is spent with Hadrian saying stuff below
*Something must be wrong! *
Seventy seconds to run one test! Then to accept the new output is another 70 seconds. If I have a couple of dozen test-output wibbles to fix, that adds up to a lot of time.
Simon
]0;Starting...(/home/simonpj/code/HEAD-31) ]0;Running for 5s [119/38037], predicted 264m08s (0%)(/home/simonpj/code/HEAD-31) ]0;Running for 10s [8853/38038], predicted 36m59s (0%)(/home/simonpj/code/HEAD-31) ]0;Running for 15s [12109/38038], predicted 9m46s (2%)(/home/simonpj/code/HEAD-31) ]0;Running for 20s [13119/38038], predicted 7m03s (4%)(/home/simonpj/code/HEAD-31) ]0;Running for 25s [13879/38038], predicted 7m19s (5%)(/home/simonpj/code/HEAD-31) ]0;Running for 30s [14485/38038], predicted 6m58s (6%)(/home/simonpj/code/HEAD-31) ]0;Running for 35s [15338/38038], predicted 6m06s (8%)(/home/simonpj/code/HEAD-31) ]0;Running for 40s [16434/38038], predicted 2m56s (17%)(/home/simonpj/code/HEAD-31) ]0;Running for 45s [20588/38038], predicted 3m32s (16%)(/home/simonpj/code/HEAD-31) ]0;Running for 50s [26354/38038], predicted 63m21s (1%)(/home/simonpj/code/HEAD-31) ]0;Running for 55s [30040/38038], predicted 49m36s (1%)(/home/simonpj/code/HEAD-31) ]0;Running for 1m00s [35123/38038], predicted 19m59s (4%)(/home/simonpj/code/HEAD-31) # ghc _______________________________________________ ghc-devs mailing list -- ghc-devs@haskell.org To unsubscribe send an email to ghc-devs-leave@haskell.org
-- brandon s allbery kf8nh allbery.b@gmail.com
_______________________________________________ ghc-devs mailing list --ghc-devs@haskell.org To unsubscribe send an email toghc-devs-leave@haskell.org
participants (3)
-
Andreas Klebinger -
Brandon Allbery -
Simon Peyton Jones