Euphoria
Ticket #685:
minor changes are needed for GNU Makefile for MING targets to work properly
-
Reported by
SDPringle
Jun 28, 2011
Due to the ways Windows differs from UNIX and due to the fact that there have been various changes that hiterto not tested on MinGW there are some targets that are broken on MinGW. Changes are needed to rectify this.
Details
1. Comment by SDPringle
Jun 28, 2011
changeset 36d933c117e7 makes the configure script correctly make eu.cfg. see ticket 683: for changes within EUPHORIA's internal build system.
2. Comment by SDPringle
Jun 28, 2011
See: hg:euphoria/rev/3bdc21ab59b5
changeset: 5003:3bdc21ab59b5 branch: 4.0 tag: tip user: Shawn Pringle <shawn.pringle@gmail.com> date: Tue Jun 28 14:20:23 2011 -0300 files: source/Makefile.gnu description:
- for ticket:685
- targets tested: tools, test
3. Comment by SDPringle
Jun 28, 2011
At first coverage seems to work, but later the include 'eucoverage.ex' is run with '-i c;C:\dir\file' from somewhere.
4. Comment by mattlewis
Jun 29, 2011
When running the coverage target, eucoverage.ex is run by eutest in process_coverage().
Its command line is determined by the makefile
-coverage-pp "$(EXE) -i $(CYPTRUNKDIR)/include $(CYPTRUNKDIR)/bin/eucoverage.ex"
So I guess CYPTRUNKDIR is messed up somehow.
5. Comment by ne1uno
Jul 01, 2011
make test still fails for me from msys rxvt shell.
with the changes to build/eu.cfg to normal paths with forward slashes. I can make test from a windows command prompt w/o editing eu.cfg. but now it exits after one of the http tests finishes. probably because I was offline?
You may now run C:\c\c\make_all\eu4\eu4m\tests\t_net_http.exe
failed: get_url 1, expected: TRUE but got: FALSE
failed: get_url 2, expected: TRUE but got: FALSE
failed: get_url 3, expected: TRUE but got: FALSE
failed: get_url post 1, expected: TRUE but got: FALSE
failed: get_url post 2, expected: TRUE but got: FALSE
failed: get_url post 3, expected: TRUE but got: FALSE
failed: get_url post 4, expected: TRUE but got: FALSE
failed: get_url post 5, expected: TRUE but got: FALSE
failed: get_url post 6, expected: TRUE but got: FALSE
failed: get_url post 7, expected: TRUE but got: FALSE
failed: get_url post 8, expected: TRUE but got: FALSE
failed: multipart form file upload, expected: TRUE but got: FALSE
12 tests run, 0 passed, 12 failed, 0% success
make: *** [test] Error 5
there is also a 1gig? memory request that fails in t_text.e earlier
include\std\serialize.e:95 in function deserialize_file()
Your program has run out of memory.
One moment please...
68 tests run, 67 passed, 1 failed, 99% success
couldn't alloc 1006381096 bytes
FAILURE: t_text.e EUPHORIA error with status 1
haven't yet confirmed the tests run ok w/watcom built tools yet.
6. Comment by ne1uno
Jul 01, 2011
not saying it's worth debugging why make test fails from the msys rxvt shell since you can run the tests from a windows command prompt and probably other window shells ok now.
msys aspires to run configure make and little else. many projects won't entertain bugs related to msys.
but I did want to add where it fails in case it is a simple fix or obvious to someone else why.
a bunch of tests run normally, then it starts to go bad here
interpreting t_c_map_large_bad_init.e:
/include\std\map.e:751 in procedure put()
Inappropriate initial operation given to map.e:put()
... called from \t_c_map_large_bad_init.e:7
--> See ex.err
interpreting t_c_map_large_bad_next.e:
/include\std\map.e:741 in procedure put()
Unknown operation given to map.e:put()
... called from \t_c_map_large_bad_next.e:8
--> See ex.err
---------
a bunch more interpreted tests pass, fail or crash then
---------
interpreting t_c_safe_badpoke.e:
Using Debug Version of machine.e
include\std\safe.e:263 in procedure die()
BAD POKE ADDRESS!!!! 3735928559 (#DEADBEEF)
... called from include\std\safe.e:458 in procedure poke()
... called from \t_c_safe_badpoke.e:10
--> See ex.err
interpreting t_c_safe_c_func_nodep.e:
Using Debug Version of machine.e
include\std\dll.e:374 in function define_c_func()
A C function is being defined from Non-executable memory.
... called from \t_c_safe_c_func_nodep.e:18
--> See ex.err
interpreting t_c_safe_free_allocate_protect.e:
make: *** [coverage] Error 5
7. Comment by SDPringle
Jul 06, 2011
See: hg:euphoria/rev/cb79f84e0276
changeset: 5047:cb79f84e0276 branch: 4.0 parent: 5041:7336baef56ca user: Shawn Pringle <shawn.pringle@gmail.com> date: Sat Jul 02 12:11:07 2011 -0300 files: source/configure description:
- fix ticket 685
- Now uses -W switch in Mingw for pwd, which helps if you do 'cd;cd TRUNKDIR'
- hg not passed to which. This could cause problems if run outside of the shell environment and it doesn't change the behavior.
8. Comment by SDPringle
Aug 04, 2011
See: hg:euphoria/rev/37a39c7f43eb
changeset: 5090:37a39c7f43eb branch: sse2 tag: tip user: Shawn Pringle <shawn.pringle@gmail.com> date: Thu Aug 04 16:49:45 2011 -0300 files: demo/bench/math_ops.ex source/Makefile.wat source/configure.bat description:
- changes to Makefile and configure to make building smoother
- fix ticket 685
9. Comment by SDPringle
Sep 16, 2011
The translator dies on account of specifying things in both eu.cfg and at the command line. We shouldn't have a eu.cfg in the source directory; only in the build directory. We can move common translator options to this eu.cfg file in the build directory.
10. Comment by mattlewis
Sep 16, 2011
I think we need the eu.cfg in the source directory, too. What is failing, specifically? This is more likely a bug in how we're handling configuration options.
11. Comment by SDPringle
Sep 18, 2011
See: hg:euphoria/rev/96e909de9604
changeset: 5156:96e909de9604 branch: 4.0 tag: tip parent: 5154:e4b0bd393ab7 user: Shawn Pringle <shawn.pringle@gmail.com> date: Sun Sep 18 22:34:22 2011 -0300 files: source/Makefile.gnu source/buildsys.e source/configure tests/t_filesys.e description:
- fixes ticket 685
- moved switches from Makefile into $BUILDDIR/eu.cfg
- prevented creating eu.cfg in source directory (configure / bash)
- adjusted translator to always use forward slashes in creating makefiles for GCC even in Windows when adjusted to command line passing: The environment is always bash.
- adjusted t_filesys.e to get home directory even on Windows and to use this value if it exists. HOME is set for MINGW.
12. Comment by SDPringle
Nov 12, 2011
See: hg:euphoria/rev/a4566a32c7d2
changeset: 5229:a4566a32c7d2 branch: alternative_literals tag: tip user: Shawn Pringle <shawn.pringle@gmail.com> date: Sat Nov 12 17:58:08 2011 -0300 files: source/configure description:
- ticket 685... ARCH setting in configure
GitHub