MinGW - Minimalist GNU for Windows

Fabian Greffrath wrote:
> Am 28.11.2011 15:40, schrieb Earnie:
>> For bash-4.x using the current MSYS or the stuff Ralph did.  Note that
>> what I captured into the git repository and patches waiting another
>> review did not include the entire set of Ralph's work, it was a bit over
>> done, nor does it include all that needs to be modified.
> 
> I was asking about the stuff Ralph did. I mean, rebasing MSYS on 
> Cygwin 1.9 will certainly involve more work than taking said Cygwin 
> version and doing s/cygwin/msys/g. What I was asking for was a "list" 
> of differences between Cygwin and MSYS that need to get 
> ported/implemented.
> 

Once the runtime works, the rest should be a matter of rebuild with the
MSYS compiler.

>> Fine.  You'll either need one of two things or both.  Find what you need
>> to change in bash or find what you need to change in MSYS runtime.
> 
> I guess if the rebasing of MSYS on currentl Cygwin succeeds, this will 
> be a lot easier, i.e. I hope we can take the patches from Cygwin for 
> most packages and rebuild them for the MSYS runtime - modulo the "list 
> of things in which MSYS differs from Cygwin". ;)
> 

Yes, but if you're interested in just bash-4.x then you might just give
a try to the configure && make to find what breaks.  There are some MSYS
specific changes such as ``pwd -W'' that do need to be included.

-- 
Earnie
-- https://sites.google.com/site/earnieboyd/

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

I can tell i allready built bash 4 with msys with parts of the old
patch but it segfaults a lot :/.

Im still on the beginners stage in regards to debugging with strace
and gdb which makes it guesswork to find out what part fails.

One qualified guess would be the built in readline library which i had
numerous problems getting to work correctly on later versions (noticed
this fault with my experiment with a native msys python which crashed
in the same way). I use an older readline library now which works.
The readline i used before was the newest version compiled by myself
with ncurses support, unfortunatly bash wants the wide version which
cannot be built with msys due to missing unicode support.

If we get it to work it will be a lobotomized version anyway :P so
maybe better to wait untill the new core is ready (just a thought).

Ralph
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (MingW32)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iQEcBAEBAgAGBQJO1JTHAAoJEIjGvG7Y4HU8QbAH/ROA+XbQsKFG1NVZ94eRiHee
HdP5HOqfIEbCltxxFFsaEFhG7ebTkqI31sms9PhJTtLvt6z0/LBvVJng3NLbrDgF
Qhg6mmujlvRxK9jkTje2tENxPiV9aACxAwt25jBBDaCN4wKjhHcXoiwL2xliWfnQ
4h3STR+sdq5SWBa24n0IFhIsybm31Hbo4gxCWbtXJtgegFbNSAv1fJo7gWww6tMp
Km8AEO1GHUPxU2J5k7ztCwdmzYHcqmmmqFtZljgIgAi1P3srtnGRViq/AgWrGqso
7Z5+7gwMtZBMQYSdF2ApEynZOFf+QBIkX2pgy/PcPiQterwFfRuLceJp8+NiFH0=
=Bl2L
-----END PGP SIGNATURE-----

Am 28.11.2011 21:07, schrieb Earnie:
> Yes, but if you're interested in just bash-4.x then you might just give
> a try to the configure&&  make to find what breaks.  There are some MSYS
> specific changes such as ``pwd -W'' that do need to be included.

Fine, that's exactly the stuff I was asking for: What's the exact 
difference between MSYS and Cygwin apart from 's/cygwin/msys/g' in the 
source code?

I can tell a few of them from my head, e.g. different mount points for 
the hard disk partitions (/cygdrive/c/ vs. /c/) and I think MSYS is 
more tolerant with respect to CRLF line endings in its config files.

What else? Is there some kind of recipe that tells what has to get 
ported or implemented in order to turn Cygwin into MSYS?

  - Fabian

Am 01.12.2011 14:14, schrieb Fabian Greffrath:
> Fine, that's exactly the stuff I was asking for: What's the exact
> difference between MSYS and Cygwin apart from 's/cygwin/msys/g' in the
> source code?

Related question: On the mingw.org page I learn that MSYS is "a 
light-weight fork of Cygwin-1.3".

How is "light-weight" defined if MSYS is more or less a renamed code 
copy of Cygwin with some specific changes?

Fabian Greffrath wrote:
> Am 28.11.2011 21:07, schrieb Earnie:
>> Yes, but if you're interested in just bash-4.x then you might just give
>> a try to the configure&&  make to find what breaks.  There are some MSYS
>> specific changes such as ``pwd -W'' that do need to be included.
> 
> Fine, that's exactly the stuff I was asking for: What's the exact 
> difference between MSYS and Cygwin apart from 's/cygwin/msys/g' in the 
> source code?
> 

I tend to forget everything required.  But if interested I would compare
the two sources.

> I can tell a few of them from my head, e.g. different mount points for 
> the hard disk partitions (/cygdrive/c/ vs. /c/) and I think MSYS is 
> more tolerant with respect to CRLF line endings in its config files.
> 

MSYS runtime handles the /cygdrive thing.  Actually just sets the
default to / and doesn't provide the user a chance to change it.  And
MSYS is less tolerant of CRLF line endings, i.e. you must have LF only
line endings.

> What else? Is there some kind of recipe that tells what has to get 
> ported or implemented in order to turn Cygwin into MSYS?

Again, compare the 3.1.17 sources between Cygwin and MSYS, I don't know
what else.

-- 
Earnie
-- https://sites.google.com/site/earnieboyd/

Fabian Greffrath wrote:
> Am 01.12.2011 14:14, schrieb Fabian Greffrath:
>> Fine, that's exactly the stuff I was asking for: What's the exact
>> difference between MSYS and Cygwin apart from 's/cygwin/msys/g' in the
>> source code?
> 
> Related question: On the mingw.org page I learn that MSYS is "a 
> light-weight fork of Cygwin-1.3".
> 
> How is "light-weight" defined if MSYS is more or less a renamed code 
> copy of Cygwin with some specific changes?
> 

Much of the Cygwin source still resides untouched (there is a lot of
source code in Cygwin only a small portion of that code was changed).

-- 
Earnie
-- https://sites.google.com/site/earnieboyd/

Am 01.12.2011 14:23, schrieb Earnie:
> MSYS runtime handles the /cygdrive thing.  Actually just sets the
> default to / and doesn't provide the user a chance to change it.  And
> MSYS is less tolerant of CRLF line endings, i.e. you must have LF only
> line endings.
>> [...]
> Again, compare the 3.1.17 sources between Cygwin and MSYS, I don't know
> what else.

Sorry, I was actually talking about the runtime, not about bash.

Fabian Greffrath wrote:
> Am 01.12.2011 14:23, schrieb Earnie:
>> MSYS runtime handles the /cygdrive thing.  Actually just sets the
>> default to / and doesn't provide the user a chance to change it.  And
>> MSYS is less tolerant of CRLF line endings, i.e. you must have LF only
>> line endings.
>>> [...]
>> Again, compare the 3.1.17 sources between Cygwin and MSYS, I don't know
>> what else.
> 
> Sorry, I was actually talking about the runtime, not about bash.

Thread hijack confusion.  Start a new thread.

-- 
Earnie
-- https://sites.google.com/site/earnieboyd/

On 12/1/2011 11:24 AM, Fabian Greffrath wrote:
> Am 01.12.2011 14:14, schrieb Fabian Greffrath:

> Related question: On the mingw.org page I learn that MSYS is "a
> light-weight fork of Cygwin-1.3".
>
> How is "light-weight" defined if MSYS is more or less a renamed code
> copy of Cygwin with some specific changes?

MSYS disables symlink emulation and the mapping of Windows Access 
Control Lists to POSIX permissions.

Cesar

Fabian Greffrath skrev 2011-12-01 14:24:
> Am 01.12.2011 14:14, schrieb Fabian Greffrath:
>> Fine, that's exactly the stuff I was asking for: What's the exact
>> difference between MSYS and Cygwin apart from 's/cygwin/msys/g' in the
>> source code?
> 
> Related question: On the mingw.org page I learn that MSYS is "a 
> light-weight fork of Cygwin-1.3".
> 
> How is "light-weight" defined if MSYS is more or less a renamed code 
> copy of Cygwin with some specific changes?

MSYS does not fiddle with the process error mode (SetErrorMode) in the
same way Cygwin does, which means that you get popup dialogs when you
are trying to run programs that are missing DLLs etc (instead of a
more subtle non-zero exit status, only). But I don't know if that is
an MSYS feature to keep when doing a new fork?

Cheers,
Peter

Am 02.12.2011 09:04, schrieb Peter Rosin:
> MSYS does not fiddle with the process error mode (SetErrorMode) in the
> same way Cygwin does, which means that you get popup dialogs when you
> are trying to run programs that are missing DLLs etc (instead of a
> more subtle non-zero exit status, only). But I don't know if that is
> an MSYS feature to keep when doing a new fork?

Thanks for pointing this out! Unfortunately it is close to impossible 
to identify these subtle differences merely from reading the msys-core 
source code. I remember that Teemu Nätkinniemi once posted a patch 
against Cygwin-1.3 that reflected the first MSYS revision, but this 
specific commit has been checked in ten years ago and the msys-core 
code has developed severly since then.

It would be ideal if msys-core could be handled a some kind of static 
patch against Cygwin, so one can always track the changes between 
"upstream" (i.e. Cygwin) and MSYS and rebase on new upstream releases 
if necessary. This literally screams for git. ;)

  - Fabian

Am 02.12.2011 09:04, schrieb Peter Rosin:
> MSYS does not fiddle with the process error mode (SetErrorMode) in the
> same way Cygwin does, which means that you get popup dialogs when you
> are trying to run programs that are missing DLLs etc (instead of a
> more subtle non-zero exit status, only). But I don't know if that is
> an MSYS feature to keep when doing a new fork?

Thanks for pointing this out! Unfortunately it is close to impossible 
to identify these subtle differences merely from reading the msys-core 
source code. I remember that Teemu Nätkinniemi once posted a patch 
against Cygwin-1.3 that reflected the first MSYS revision, but this 
specific commit has been checked in ten years ago and the msys-core 
code has developed severly since then.

It would be ideal if msys-core could be handled a some kind of static 
patch against Cygwin, so one can always track the changes between 
"upstream" (i.e. Cygwin) and MSYS and rebase on new upstream releases 
if necessary. This literally screams for git. ;)

  - Fabian

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

The new fork (while closed atm) is on git and will probably stay there
as i think earnie has some plans in regard to msysgit.

And i agree that we should keep a static patchset so users easily can
follow changes in either source. Unfortunatly for earnie i was a bit
to litteral when i started my port from cygwin-1.9.1 i replaced
instances of cygwin with msys where i should not have (like
documentation etc.) and were backporting from there. Hopefully wont
take to long but i been a bit unlucky lately with my health (just
returned from surgery having my gallbladder removed) and still recovering.

Ralph

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (MingW32)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iQEcBAEBAgAGBQJO2KqmAAoJEIjGvG7Y4HU8h0QIAJ/qp9M0fggQnu7vhiZYilTp
vczgXcDPH8MzA+w1aE/UiYcQPkAgLJvP/JGNvgsoMGQ1jH+9kPy4ry8mQh4WDJq8
XIAu59IR7mumBpitCvcWHoHRHxmpKl/Y8Ai10AAO8wjrjiZlVX4L2y+LqIsnewAd
m3iXQ7tBwwneQGI1IEHdD4cwPDO/1bUFdUhIZfOJj+4cDrNg/nQy7DkzrAsJ1dAC
f0rloXSBK80xQgEqH4GFHgUbiLDj9OGK14D5BN9S/tcgxS8RBry1dR9rfJLU7L7R
HpQ3EHs13zlZfyxay3pNMVQC2aUyIe//IZQLsEFD0kwnTIcO7RX8xRZ2ZXBVAZ0=
=q3xq
-----END PGP SIGNATURE-----

Peter Rosin wrote:
> Fabian Greffrath skrev 2011-12-01 14:24:
>> Am 01.12.2011 14:14, schrieb Fabian Greffrath:
>>> Fine, that's exactly the stuff I was asking for: What's the exact
>>> difference between MSYS and Cygwin apart from 's/cygwin/msys/g' in the
>>> source code?
>>
>> Related question: On the mingw.org page I learn that MSYS is "a 
>> light-weight fork of Cygwin-1.3".
>>
>> How is "light-weight" defined if MSYS is more or less a renamed code 
>> copy of Cygwin with some specific changes?
> 
> MSYS does not fiddle with the process error mode (SetErrorMode) in the
> same way Cygwin does, which means that you get popup dialogs when you
> are trying to run programs that are missing DLLs etc (instead of a
> more subtle non-zero exit status, only). But I don't know if that is
> an MSYS feature to keep when doing a new fork?

Are you comparing apples to oranges?  I.E. is your comparison of what
Cygwin does based on the 1.3.3 version of Cygwin or is it based on what
is delivered today as Cygwin?  I don't believe this has any relevance to
changes made to Cygwin for MSYS.

-- 
Earnie
-- https://sites.google.com/site/earnieboyd/

Fabian Greffrath wrote:
> It would be ideal if msys-core could be handled a some kind of static 
> patch against Cygwin, so one can always track the changes between 
> "upstream" (i.e. Cygwin) and MSYS and rebase on new upstream releases 
> if necessary. This literally screams for git. ;)

Yes, it certainly would be nice.  Have you looked at
http://bit.ly/tO0oBP to see what I've taken from Ralph.

-- 
Earnie
-- https://sites.google.com/site/earnieboyd/

Am 02.12.2011 14:54, schrieb Earnie:
> Yes, it certainly would be nice.  Have you looked at
> http://bit.ly/tO0oBP to see what I've taken from Ralph.

Yes, the second paragraph on that page is exactly what I had in mind. 
I don't see anything from Ralph there, though.

BTW, what is Wosix? It sounds interesting, is there anything 
implemented, yet? What will be the difference to msys-*.dll?

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Nothing there from me yet besides the things earnie ripped from my
initial patch. I hope to get more involved again now that im back from
surgery :) still recovering but im allready looking into some things
so maybe soon.

Ralph
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (MingW32)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iQEcBAEBAgAGBQJO2ONwAAoJEIjGvG7Y4HU8wbwH/3tx72zflkKM/tZvWgCMF7Z9
MwVSW8f5gubgDFjRfrs5oR/XK9ypMDLgYuc0vvXc4VapVz6CNRVAUODRl5Yz8NBv
UUv1pXKOAlt0QeO/8zHBbqltI069sB7cZqUsJHs2rAAlK2VqGKr9M4tia/9k5Tgc
HiH+cjkWkAY6JHjTmg6jSARA8MDU+TU/dCK/FyQS4geZVwr3WVznWjxtwcalcfw6
4H50yFmqIxY0n2rebVrdWjpA58rdxzpza7kKOrcBhCpi8+kxEFhFtWBjiOm2CmUi
7nA3AhicrJR3AYH9DpoM8Oqsj2WemxbS62l3PiYCUOZRleZ1kgMWsp1DzbMX7IU=
=cgMo
-----END PGP SIGNATURE-----

Am 02.12.2011 15:40, schrieb Ralph Engels:
> Nothing there from me yet besides the things earnie ripped from my
> initial patch. I hope to get more involved again now that im back from
> surgery :) still recovering but im allready looking into some things
> so maybe soon.

Get well soon, I wish you all the best!

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Thanks :) should help a lot with the pain i been having (stones in my
gallbladder removed... well actually they removed the whole thing)
something i wouldnt even wish for my worst enemy to endure ouch extreme.

Ralph
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (MingW32)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iQEcBAEBAgAGBQJO2Oo5AAoJEIjGvG7Y4HU8aTUH/0VUZI/TLkY/FSxNxFgT4HHq
NbnJHaKqbtZMj3wwZ4hy4epX7E/OxKLi/TMhcw9o27ApzjHpHLJqurOh9Y2x8Rgp
573UWPIKhkNuGrS45TR4R6nROkJVLGzLIxaEwUmMPLZmGzSgClYs6ijZnzF13Z77
2wBrQK8+anSjrsvy+g7Brhpgt6vZlFztmCflkbLMZZC0AZjTDh4O0BtHTQVDpIUq
LAQFuZABS4Sz7Cs+1w7DOqirWDqC6aO/LbbKjh3zHlCUdx0PT7ZcdUyeXDuGTwf6
zryp6zS5CArIIuIly0uAuR7IIj1oTTkEOYIf5e6sjByuCqhVB9UiWluApAsC+R8=
=P24n
-----END PGP SIGNATURE-----

Earnie skrev 2011-12-02 14:50:
> Peter Rosin wrote:
>> Fabian Greffrath skrev 2011-12-01 14:24:
>>> Am 01.12.2011 14:14, schrieb Fabian Greffrath:
>>>> Fine, that's exactly the stuff I was asking for: What's the exact
>>>> difference between MSYS and Cygwin apart from 's/cygwin/msys/g' in the
>>>> source code?
>>>
>>> Related question: On the mingw.org page I learn that MSYS is "a 
>>> light-weight fork of Cygwin-1.3".
>>>
>>> How is "light-weight" defined if MSYS is more or less a renamed code 
>>> copy of Cygwin with some specific changes?
>>
>> MSYS does not fiddle with the process error mode (SetErrorMode) in the
>> same way Cygwin does, which means that you get popup dialogs when you
>> are trying to run programs that are missing DLLs etc (instead of a
>> more subtle non-zero exit status, only). But I don't know if that is
>> an MSYS feature to keep when doing a new fork?
> 
> Are you comparing apples to oranges?

Certainly, because MSYS != Cygwin, and I also said that I didn't know
if that is an MSYS feature that it is desirable to keep so I thought
that was obvious.

>                                       I.E. is your comparison of what
> Cygwin does based on the 1.3.3 version of Cygwin or is it based on what
> is delivered today as Cygwin?  I don't believe this has any relevance to
> changes made to Cygwin for MSYS.

I have no knowledge of how Cygwin 1.3.3 behaved, I'm just stating a
current difference that I know about. I also think the popup will
not show up when MSYS runs another MSYS process, i.e. only when
running a "pure" win32 process (Cygwin will never popup error mode
induced dialogs), so it seems like a quite deliberate feature. The
feature is also in the win32-friendly category, which makes it more
likely to have an MSYS origin than a Cygwin origin. But again, I
don't know the origin of the feature.

Cheers,
Peter

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Msys today != Cygwin today allthough the oldest version of cygwin i
used was 1.5 its much closer to what msys is today - ofc the msys
relevant parts. Major parts changed in the cygwin code as of today
include unicode support while the old fork Msys used was pure ansi.
Which is also a reason why we want to start a new fork because most
new tools which we cant get win32 counterparts for today rely on that.
Much more to it than the above though but its one of the main reasons.

I believe the win32 friendly error mode was part of the original
changeset but im not 100% certain so if you can get hold on the old
source for 1.3.3 you could compare it to the msys source that should
give you some answers.

Sincerely,
Ralph
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (MingW32)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iQEcBAEBAgAGBQJO2QSfAAoJEIjGvG7Y4HU8xTAH/0d81bEwvF+ZlyvzhlM/wnyP
WGdS2FV+efTxEjJQqz3iHfxP2MJz5IwD11EwFH7kqYZORR/Cm6COisdgjoVGI19m
Zrp3KO8nOk5BXDhRaLrXI00MRUQ7wtNuc/JamPgix1IGka2BbnB7pwECPsyXeVDm
R65gHCA14OiRHDzDsjDUTc/z4wNIEVmNoxpjR82cAhXUF5MGZmYZn0wJaX3hIG86
aEv7dosrQ1Y8sF1KXhOuBmbYs1t68NdEaccQR2cILUaqASS0+dKFFvtfk1LlEONF
6Xj/U12fcWq/MCcgydr5YsT0pcR14l1Zf93TKk6KSBh10tOop8IZSFZtSJs3NKw=
=qUyl
-----END PGP SIGNATURE-----

Peter Rosin wrote:

> 
> I have no knowledge of how Cygwin 1.3.3 behaved, I'm just stating a
> current difference that I know about. I also think the popup will
> not show up when MSYS runs another MSYS process, i.e. only when
> running a "pure" win32 process (Cygwin will never popup error mode
> induced dialogs), so it seems like a quite deliberate feature. The
> feature is also in the win32-friendly category, which makes it more
> likely to have an MSYS origin than a Cygwin origin. But again, I
> don't know the origin of the feature.
> 

Fabian was asking what needed to change in the current based on what was
changed on the previous.  Stating the slices of a new breed as compared
to that of the old breed isn't what was asked which is why I asked you
what I did.  It's quite obvious the behavior between the previous and
new are different.

-- 
Earnie
-- https://sites.google.com/site/earnieboyd/

Fabian Greffrath wrote:
> BTW, what is Wosix? It sounds interesting, is there anything 
> implemented, yet? What will be the difference to msys-*.dll?

An idea not yet implemented and may never be.  If you want to know more
open a ticket on the project and ask there.

-- 
Earnie
-- https://sites.google.com/site/earnieboyd/

Fabian Greffrath wrote:
> Am 02.12.2011 14:54, schrieb Earnie:
>> Yes, it certainly would be nice.  Have you looked at
>> http://bit.ly/tO0oBP to see what I've taken from Ralph.
> 
> Yes, the second paragraph on that page is exactly what I had in mind. 
> I don't see anything from Ralph there, though.

The repositories already exist but public access has been removed.
Email me privately if you want developer access.

-- 
Earnie
-- https://sites.google.com/site/earnieboyd/