#14626: No need to enter a scrutinised value -------------------------------------+------------------------------------- Reporter: heisenbug | Owner: heisenbug Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 8.2.2 Resolution: | Keywords: performance Operating System: Unknown/Multiple | Architecture: | Unknown/Multiple Type of failure: None/Unknown | Test Case: Blocked By: | Blocking: Related Tickets: #13861 | Differential Rev(s): Wiki Page: | -------------------------------------+------------------------------------- Comment (by heisenbug): Replying to [comment:10 simonpj]:
The names of the variables should not make any difference! That's bizarre.
Of course it is not the names, ''per se''! Rather, one of the names starting with `ds` is misbehaving in a specific way, that trips over `stage2`. I built a `quick` compiler with the given revision, that is `stage2` compiled with `-O0`, and it behaves well. Testsuite passes without any strangeness. Then I bisected the sources in `compiler/*` and isolated these two (mutually dependent) sources, which when built with `-O1` (and all others with `-O0`) break stage 2. Here is what I did (for the record): {{{#!sh nice make V=1 GhcStage2HcOpts="-O0 -g" -j8 rm compiler/stage2/build/StgCmmBind.* nice make V=1 GhcStage2HcOpts="-O1 -g" -j8 ./compiler/stage2/build/StgCmmBind.hi nice make V=1 GhcStage2HcOpts="-O0 -g" -j8 }}} I'll have a look which differences occur here between `-O0` and `-O1`. Both interesting sources are >500 LOC, so it won't be trivial :-)
Guessing is usually fruitless; you need data.
Did you add that assertionn I suggested?
As I say, a stage2 compiler is a huge program. I urge you to first
build the libraries and compiler without the change; then switch the change on and run the testsuite. Any bugs must be in those little programs. Then nofib.
-- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/14626#comment:11 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler