Hi Edward, I commented in the phab. Btw., this looks like a copy-paste error: git show 6338a1cc commit 6338a1cc6df2c7fd8a62eeb4c5240dd90ee74a6c Author: Edward Z. Yang <ezyang@cs.stanford.edu> Date: Sun Oct 11 13:41:20 2015 -0700 Rename PACKAGE_KEY and LIB_NAME in build system. Signed-off-by: Edward Z. Yang <ezyang@cs.stanford.edu> diff --git a/compiler/ghc.mk b/compiler/ghc.mk index 8172ca6..fc9e891 100644 --- a/compiler/ghc.mk +++ b/compiler/ghc.mk @@ -441,11 +441,11 @@ ifeq "$(compiler_stage1_VERSION_MUNGED)" "YES" compiler_stage1_MUNGED_VERSION = $(subst .$(ProjectPatchLevel),,$(ProjectVersion)) define compiler_PACKAGE_MAGIC compiler_stage1_VERSION = $(compiler_stage1_MUNGED_VERSION) -compiler_stage1_PACKAGE_KEY = $(subst .$(ProjectPatchLevel),,$(compiler_stage1_PACKAGE_KEY)) -compiler_stage1_LIB_NAME = $(subst .$(ProjectPatchLevel),,$(compiler_stage1_LIB_NAME)) +compiler_stage1_COMPONENT_ID = $(subst .$(ProjectPatchLevel),,$(compiler_stage1_COMPONENT_ID)) +compiler_stage1_COMPONENT_ID = $(subst .$(ProjectPatchLevel),,$(compiler_stage1_COMPONENT_ID)) endef Cheers, Gabor On 10/22/15, Edward Z. Yang <ezyang@mit.edu> wrote:
Phab here: https://phabricator.haskell.org/D1355
Excerpts from Edward Z. Yang's message of 2015-10-22 13:09:25 -0700:
Scratch that, I managed to reproduced. (As you said, it occurs only when you do the bindist.) I'll debug.
Edward
Excerpts from Gabor Greif's message of 2015-10-21 23:47:37 -0700:
I did not use an inplace-stage2 but had a 'make install' before and did a boot/reconfigure.
Not sure whether this detail is relevant.
How can I debug this problem? What does the error message say precisely?
Cheers,
Gabor
On 10/22/15, Edward Z. Yang <ezyang@mit.edu> wrote:
So... I can't reproduce this. I validated a copy of GHC HEAD, and then used the inplace ghc-stage2 to build another using ./configure --with-ghc=... and it built fine.
Maybe, do you have some more details?
Edward
Excerpts from Edward Z. Yang's message of 2015-10-21 11:44:26 -0700:
This is likely my fault, w.r.t. to the recent Cabal submodule update. I'm a bit confused why there are things in the DB that don't have ABI hashes though...
Edward
Excerpts from Gabor Greif's message of 2015-10-21 03:51:02 -0700:
Hi devs,
just a heads-up (pun intended...)
(After updating to TOT, boot and configure with a freshly installed TOT ghc as bootstrap compiler. Then 'make clean'.)
Running 'make' gives:
... Creating includes/ghcplatform.h... Done. "rm" -f utils/hsc2hs/dist/build/.depend.haskell.tmp "/home/ggreif/bin/ghc" -M -static -H64m -O0 -fasm -package-db libraries/bootstrapping.conf -hide-all-packages -i -iutils/hsc2hs/. -iutils/hsc2hs/dist/build -iutils/hsc2hs/dist/build/autogen -Iutils/hsc2hs/dist/build -Iutils/hsc2hs/dist/build/autogen -optP-include -optPutils/hsc2hs/dist/build/autogen/cabal_macros.h -package-id base-4.8.2.0 -package-id containers-0.5.6.2 -package-id directory-1.2.2.0 -package-id filepath-1.4.0.0 -package-id process-1.2.3.0 -XHaskell2010 -no-user-package-db -rtsopts -odir utils/hsc2hs/dist/build -hidir utils/hsc2hs/dist/build -stubdir utils/hsc2hs/dist/build -dep-makefile utils/hsc2hs/dist/build/.depend.haskell.tmp -dep-suffix "" -include-pkg-deps utils/hsc2hs/./Main.hs utils/hsc2hs/./C.hs utils/hsc2hs/./Common.hs utils/hsc2hs/./CrossCodegen.hs utils/hsc2hs/./DirectCodegen.hs utils/hsc2hs/./Flags.hs utils/hsc2hs/./HSCParser.hs utils/hsc2hs/./UtilsCodegen.hs <command line>: package db: duplicate packages with incompatible ABIs: binary-0.7.5.0 has ABIs: , c28f822c21e75eb270eca0870e42aaac Cabal-1.23.0.0 has ABIs: , a0a6af8f1dd909f2ee719ddb2cd65779 hpc-0.6.0.2 has ABIs: , 3434375974d4bc6d14952a90ec97d0c4 ghc-boot-0.0.0.0 has ABIs: , f422fc19421064f81c42815f25a11e6e hoopl-3.10.0.2 has ABIs: , 8968e2731ff8d529c37c048d65e94bf2 transformers-0.4.3.0 has ABIs: , 812457c97c58693d7f8a813b1ba19a33 template-haskell-2.11.0.0 has ABIs: , 0ef51476100e9bdf96d1bf59696b98a1 terminfo-0.4.0.1 has ABIs: , aa24f544c0e3614d419e63fa170ac467
The only solution I could come up for now is
cd libraries ln -s /home/ggreif/lib/ghc-7.11.20151020/package.conf.d bootstrapping.conf
then resuming 'make' leads to success.
What could this be? Has somebody seen such an error?
Cheers and thanks,
Gabor