Discussion and patches for MSYS (Minimal SYStem) for MinGW

Ralph Engels | 19 Dec 2011 21:59

Picon

experimental msys package


I made some fixes to the current msys core and added a few things from
later cygwin core (no not unicode support).

then i rebuild perl-5.8.8 with threads and added the old cygwin tetex
package. I rebuilt texinfo with tex support and included docbook
support + the python and perl docbook packages to interface with tetex.
All seems to be working properly.

Also added svn git and apache (experimental) svn and git seems to work
ok but apache tends to croak a bit by starting to many threads it does
work but sometimes it crashes at startup (just try again and it will
work).

I updated perl makemaker and friends with some additional patches
(test dynamic was not working it works now).

My build system uses mingw64 so excuse the weird looking profile in /etc.
So adjust the mingw paths in my profile to fit and remove the other
paths to LLVM and MINGW64. You also need to adjust the console.xml in
/run/console32 or /run/console64 (only for 64 bit).

Added gcc-4 to the msys package use it for stubborn code that croaks
with gcc-3 i found it works quite well.

Also included autotools wrappers from 2.13 to 2.68 and automake
wrappers from 1.4 to 1.11.

If any parts of the package have interrest to the msys list drop me a
line and ill pack up my ports and patches.

(Continue reading)

Teemu Nätkinniemi | 19 Dec 2011 22:17

Picon

Re: experimental msys package

On 19.12.2011 22:59, Ralph Engels wrote:

> I made some fixes to the current msys core and added a few things from
> later cygwin core (no not unicode support).

Do you have sources and/or patches available?

Teemu

------------------------------------------------------------------------------
Write once. Port to many.
Get the SDK and tools to simplify cross-platform app development. Create 
new or port existing apps to sell to consumers worldwide. Explore the 
Intel AppUpSM program developer opportunity. appdeveloper.intel.com/join
http://p.sf.net/sfu/intel-appdev

Ralph Engels | 20 Dec 2011 01:54

Picon

Re: experimental msys package


I can pack them up about tomorrow :) which parts are you interrested
in ? only the core or the gcc patches also ? the tetex source needs a
bit more work atm i can only build it statically but it works fine.

Ralph

Teemu Nätkinniemi | 20 Dec 2011 07:51

Picon

Re: experimental msys package

On 20.12.2011 2:54, Ralph Engels wrote:

> I can pack them up about tomorrow :) which parts are you interrested
> in ? only the core or the gcc patches also ? the tetex source needs a
> bit more work atm i can only build it statically but it works fine.

Both GCC and core itself and Binutils if you have a modified version. I 
see you used 1.0.17 instead of 1.0.18 as a base for your modifications.

Teemu

------------------------------------------------------------------------------
Write once. Port to many.
Get the SDK and tools to simplify cross-platform app development. Create 
new or port existing apps to sell to consumers worldwide. Explore the 
Intel AppUpSM program developer opportunity. appdeveloper.intel.com/join
http://p.sf.net/sfu/intel-appdev

Ralph Engels | 20 Dec 2011 10:53

Picon

Re: experimental msys package


Ill pack them up.

Hmm newer noticed that there was a newer msys-core source ill have a
look if my patches still apply to it.

The gcc i used is a bit older than the one you sent me gcc-4.2.1
i had some weird problems with the 4.3.4 source (not related to newlib).

Ill write back with links to them.

Ralph

Ralph Engels | 20 Dec 2011 11:31

Picon

Re: experimental msys package


Binutils i use are 2.22.1 took less hacking than expected and seems to
work (much to my surprise) though i did compile it with gcc-4.

Some of my patches to the msyscore are based on some bugfixes from
msysgit (cannot build the core with gcc-4 otherwise).
One was a bug in ctype_.c in newlib gcc-3 would pass on that one but
gcc-4 will break with undefined references to ctype.
Another dealt with a FIXME in spawn.cc with a duplicate for loop.
And finally a fix for wildcard expansion by surrounding them with quotes.

Msys core can build with gcc-4 now but its unstable so not recomended
yet.

Ralph

Ralph Engels | 20 Dec 2011 13:08

Picon

Re: experimental msys package


gcc-4.2.1 source uploaded here.
https://sourceforge.net/projects/cbadvanced/files/Msys%20Specific/?

i might have fixed the problem building the msyscore with gcc-4 (broke
autotools before) let me run a few tests and if ok ill upload this as
well.

Ralph

Teemu Nätkinniemi | 21 Dec 2011 18:52

Picon

Re: experimental msys package

On 20.12.2011 14:08, Ralph Engels wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> gcc-4.2.1 source uploaded here.
> https://sourceforge.net/projects/cbadvanced/files/Msys%20Specific/?

I noticed that build.sh has a small typo: --disable-werror-allways 
should be --disable-werror-always

> i might have fixed the problem building the msyscore with gcc-4 (broke
> autotools before) let me run a few tests and if ok ill upload this as
> well.

That would be nice.

Teemu

------------------------------------------------------------------------------
Write once. Port to many.
Get the SDK and tools to simplify cross-platform app development. Create 
new or port existing apps to sell to consumers worldwide. Explore the 
Intel AppUpSM program developer opportunity. appdeveloper.intel.com/join
http://p.sf.net/sfu/intel-appdev

Ralph Engels | 21 Dec 2011 19:49

Picon

Re: experimental msys package


Thanks for pointing out the typo.

The uploaded msyscore-1.17-2 source on my page works with gcc-4 now.
I made a rather annoying error in the uploaded autotool wrapper
package for msys (paths where set to /opt/autotools) i reuploaded the
correct one with paths set to /usr so if you tried using it and got an
error just overwrite it with the new package.

I just finished work on the only remaining tools needed for tetex
tiff was a bitch to compile and i only succeded by doing a cross compile
from mingw to msys :S it works ok though but the build is a mess so id
rather clean it up a bit before releasing sources. Ill upload the
compiled libraries though.

Netpbm done and working and psutils also. Also added libpaper and
recompiled groff and tetex to pick up the new libraries. I can now
build mingw with full documentation in html and pdf.

This will be the last ports i do for the old runtime as im quite happy
with it now. Next ill clean up my packages and sources and convert my
build scripts to use mgwport.

Teemu Nätkinniemi | 22 Dec 2011 11:14

Picon

Re: experimental msys package

On 21.12.2011 20:49, Ralph Engels wrote:

> This will be the last ports i do for the old runtime as im quite happy
> with it now. Next ill clean up my packages and sources and convert my
> build scripts to use mgwport.

I noticed you managed to build tcl, tk and Expect so I built Dejagnu 1.5 
so we can run make check for GCC and Binutils.

The test results:

		===  Summary ===

# of expected passes		56
# of unexpected failures	2
runtest completed at Wed Dec 21 23:03:22 2011

http://teemun.x10.mx/dejagnu-1.5-msys/

Teemu

------------------------------------------------------------------------------
Write once. Port to many.
Get the SDK and tools to simplify cross-platform app development. Create 
new or port existing apps to sell to consumers worldwide. Explore the 
Intel AppUpSM program developer opportunity. appdeveloper.intel.com/join
http://p.sf.net/sfu/intel-appdev

Teemu Nätkinniemi | 22 Dec 2011 16:05

Picon

Re: experimental msys package

On 22.12.2011 12:14, Teemu Nätkinniemi wrote:
> On 21.12.2011 20:49, Ralph Engels wrote:
>
>> This will be the last ports i do for the old runtime as im quite happy
>> with it now. Next ill clean up my packages and sources and convert my
>> build scripts to use mgwport.
>
> I noticed you managed to build tcl, tk and Expect so I built Dejagnu 1.5
> so we can run make check for GCC and Binutils.

I ran make check with non-bootstrapped GCC 4.2.1 with Ralph's MSYS 
1.0.17-2 and here are the results which are similar than my earlier 
tests with different GCC versions. Most of the errors are in gcc.dg 
testsuite.

		=== gcc Summary ===

# of expected passes		28866
# of unexpected failures	13389
# of unexpected successes	27
# of expected failures		88
# of unresolved testcases	78
# of untested testcases		242
# of unsupported tests		395

Teemu

------------------------------------------------------------------------------
Write once. Port to many.
Get the SDK and tools to simplify cross-platform app development. Create

(Continue reading)

Ralph Engels | 23 Dec 2011 05:50

Picon

Re: experimental msys package


Remember to use mintty when using expect as the terminal is not
working as it should else.

I also managed to build insight for msys it uses the now ancient
version 6 gdb it works ok for msys though.

Ralph

Ralph Engels | 23 Dec 2011 05:37

Picon

Re: experimental msys package


Aye though wilson suggest we put it on the backburner as hees fixing
some of the nasty stuff with tcl/tk but feel free to use my version
until then :).

Btw dejagnu was allready in my msys update package i havent upladed it
standalone yet though.

Ralph

Teemu Nätkinniemi | 23 Dec 2011 09:26

Picon

Re: experimental msys package

On 23.12.2011 6:37, Ralph Engels wrote:

> Aye though wilson suggest we put it on the backburner as hees fixing
> some of the nasty stuff with tcl/tk but feel free to use my version
> until then :).

As he hasn't released TK yet I used your version.

> Btw dejagnu was allready in my msys update package i havent upladed it
> standalone yet though.

I am using my own MSYS installation with some of your releases.

Anyway, the next step is going to start using GCC 4.21 built MSYS and 
see how that works. Before that I am going build a bootstrapped version 
of GCC 4.21 and see if the results are improved at all although I 
seriously doubt that.

By the way, if you want to build any more recent version of GCC the heap 
size of MSYS has to be increased at least to 512 megabytes!

Teemu

Index: shared.cc
===================================================================
RCS file: /cvsroot/mingw/msys/rt/src/winsup/cygwin/shared.cc,v
retrieving revision 1.10
diff -u -r1.10 shared.cc
--- shared.cc	19 Apr 2003 01:03:48 -0000	1.10
+++ shared.cc	4 May 2011 13:50:01 -0000

(Continue reading)

Charles Wilson | 25 Dec 2011 03:52

Picon

Re: experimental msys package

On 12/22/2011 11:37 PM, Ralph Engels wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Aye though wilson suggest we put it on the backburner as hees fixing
> some of the nasty stuff with tcl/tk

I've got a build, and oddly the *test* programs seem to work fairly 
well, but wish itself (interactive or scripted) does not.  What I need 
to do is tease out all my patches into smaller, related chunks and then 
analyze them (and the behavior).  Maybe I "went too far" and by 
eliminating some of my changes I could fix the problem -- if my possible 
overzealousness caused the issue.  git-bisect is your friend (assuming 
you have more than just one giant single commit   -- in which case 
"git rebase --interactive/git-add -patch" to the rescue 
http://tomayko.com/writings/the-thing-about-git)

Or, splitting up the giant commit and analyzing the pieces may help 
convince me they are all needed, the problem lies elsewhere, and I need 
to patch even more.  Dunno yet.

But that's hard -- debugging an *interpreter* is decidedly nontrivial -- 
and I've had zero free time lately.

--
Chuck

------------------------------------------------------------------------------
Write once. Port to many.
Get the SDK and tools to simplify cross-platform app development. Create

(Continue reading)

Ralph Engels | 25 Dec 2011 04:24

Picon

Re: experimental msys package


I had a few problems with the port of wish to msys as well most
notably it fails to create a gui when run directly from a tcl script.
Weirdly enough it works ok if the tcl script uses a wrapper executable
:S (insight feks.) my guess is that the wish executable itself is
broken but the wish runtime itself works.

P.s merry christmas.

Ralph

Ralph Engels | 27 Dec 2011 17:16

Picon

Re: experimental msys package


Hmm oddity :S building gcc-4.3.4 on msys fails but 4.4.0 succedes
(well libgomp fails). But i have a working gcc-4.4.0 for msys now.

Maybe something for teemu ?.

Ralph

Teemu Nätkinniemi | 27 Dec 2011 18:45

Picon

Re: experimental msys package

On 27.12.2011 18:16, Ralph Engels wrote:

> Hmm oddity :S building gcc-4.3.4 on msys fails but 4.4.0 succedes
> (well libgomp fails). But i have a working gcc-4.4.0 for msys now.
>
> Maybe something for teemu ?.

Thanks a lot! Can you upload the sources as well?

Teemu

------------------------------------------------------------------------------
Write once. Port to many.
Get the SDK and tools to simplify cross-platform app development. Create 
new or port existing apps to sell to consumers worldwide. Explore the 
Intel AppUpSM program developer opportunity. appdeveloper.intel.com/join
http://p.sf.net/sfu/intel-appdev

Ralph Engels | 27 Dec 2011 21:33

Picon

Re: experimental msys package


Will do.
The built version is up allready (investigating why gcc-4.3.4 fails)
but ill upload the source soon as well.

Ralph

Ralph Engels | 27 Dec 2011 23:27

Picon

Re: experimental msys package


Here you go.
http://sourceforge.net/projects/cbadvanced/files/Msys%20Specific/msys-gcc-4.4.0-src.7z/download

Ralph

Ralph Engels | 31 Dec 2011 01:27

Picon

Re: experimental msys package


Ok seems to be a problem with autotools causing gcc-4.3.4 to fail
building, gcc-4.4.0 seems to work ok though but if you find anything
out of order id like to hear.

Happy new year.

Ralph

Teemu Nätkinniemi | 31 Dec 2011 04:40

Picon

Re: experimental msys package

On 31.12.2011 2:27, Ralph Engels wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Ok seems to be a problem with autotools causing gcc-4.3.4 to fail
> building, gcc-4.4.0 seems to work ok though but if you find anything
> out of order id like to hear.

How did you manage to avoid the following error with GCC 4.4.0 as EILSEQ 
shouldn't be defined in MSYS:

make[2]: Entering directory 
`/src/msys-gcc-4.4.0-src/build/i686-pc-msys/libstdc++-v3/src'
/bin/sh ../libtool --tag CXX --mode=compile 
/src/msys-gcc-4.4.0-src/build/./gcc/xgcc -shared-libgcc 
-B/src/msys-gcc-4.4.0-src/build/./gcc -nostdinc++ 
-L/src/msys-gcc-4.4.0-src/build/i686-pc-msys/libstdc++-v3/src 
-L/src/msys-gcc-4.4.0-src/build/i686-pc-msys/libstdc++-v3/src/.libs 
-B/usr/i686-pc-msys/bin/ -B/usr/i686-pc-msys/lib/ -isystem 
/usr/i686-pc-msys/include -isystem /usr/i686-pc-msys/sys-include 
-I/src/msys-gcc-4.4.0-src/build/i686-pc-msys/libstdc++-v3/include/i686-pc-msys 
-I/src/msys-gcc-4.4.0-src/build/i686-pc-msys/libstdc++-v3/include 
-I/src/msys-gcc-4.4.0-src/gcc-4.4.0/libstdc++-v3/libsupc++ 
-fno-implicit-templates -Wall -Wextra -Wwrite-strings -Wcast-qual 
-fdiagnostics-show-location=once  -ffunction-sections -fdata-sections 
-fno-function-sections -fno-data-sections -fno-unit-at-a-time 
-I/usr/include -I/usr/include/w32api    -std=gnu++0x -c 
/src/msys-gcc-4.4.0-src/gcc-4.4.0/libstdc++-v3/src/functexcept.cc
libtool: compile:  /src/msys-gcc-4.4.0-src/build/./gcc/xgcc 
-shared-libgcc -B/src/msys-gcc-4.4.0-src/build/./gcc -nostdinc++

(Continue reading)

Ralph Engels | 31 Dec 2011 04:51

Picon

Re: experimental msys package


By adding this to CPPFLAGS -DEILSEQ=138 but i took the liberty of
adding the missing define to errno.h allthough its wrong since
msyscore does not support it internally. Ill be adding support for it
later though.

same hack as the above is used in a few msys packages.

Teemu Nätkinniemi | 31 Dec 2011 05:05

Picon

Re: experimental msys package

On 31.12.2011 5:51, Ralph Engels wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> By adding this to CPPFLAGS -DEILSEQ=138 but i took the liberty of
> adding the missing define to errno.h allthough its wrong since
> msyscore does not support it internally. Ill be adding support for it
> later though.

Ok, I defined ENOENT as EILSEQ to work around this. My aim with MSYS 1.0 
has been to patch it incrementally until it gets proper wchar support 
and improved threading. The wchar support was committed on August 23rd 
2002 according to Newlib's changelog but I haven't had the time to 
actually start working on it and then again MSYS2 would fix all of the 
problems there are at the moment.

Can you upload your version of Binutils, please?

Teemu

------------------------------------------------------------------------------
Ridiculously easy VDI. With Citrix VDI-in-a-Box, you don't need a complex
infrastructure or vast IT resources to deliver seamless, secure access to
virtual desktops. With this all-in-one solution, easily deploy virtual 
desktops for less than the cost of PCs and save 60% on VDI infrastructure 
costs. Try it free! http://p.sf.net/sfu/Citrix-VDIinabox

Ralph Engels | 31 Dec 2011 07:56

Picon

Re: experimental msys package


Will do just been a bit busy :) newyear etc.

Msys2 will most likely fix that aye but it might take a while so i
dont see anything in the way for patching the old msys core :) i been
patching on it for some time now adding better threading (as you) and
some general fixes towards building with newer gcc.

Also working on some upgrades to the released doom3 source so im full
up on work heh.

Ralph

Ralph Engels | 20 Dec 2011 13:39

Picon

Re: experimental msys package


Ok msyscore seems to work fine now i uploaded it here.
https://sourceforge.net/projects/cbadvanced/files/Msys%20Specific/?

changes in the msysCORE.RELEASE_NOTES.txt

Ralph

Return

Return to gmane.comp.gnu.mingw.msys.

Search Archive

Language

Change language

Options

Current view: Threads only / Showing only 20 lines / Not hiding cited text.
Change to All messages, whole messages, or hide cited text.

Post a message
NNTP Newsgroup
Classic Gmane web interface
XML

XML

RSS Feed
List Information

About Gmane