New package behaviour of setup.ini

New package behaviour of setup.ini

[Dave Korn]

  I want to make /very/ sure of this, which is why I'm asking:

  If an entirely new package appears on the mirrors, setup.exe won't install
it automatically, even if in the past you have maybe set the entire "Devel"
category (as it happens to be in this case, but I could have said the entire
top-level "All" category) to "Install"?  Or will it?

  The experimental 'gcc4' packages I'm ready to upload are incompatible
packaging-wise with the current 'gcc' packages; different requires: lines,
different number of packages, only one common source package for the lot.

  So I don't think I can make it just a new -4.3.0-1 version of the existing
package.

  That means that anyone who wants to play with it needs to manually uninstall
their existing gcc first, and then install gcc4.  (This can of course be done
safely in a single pass of setup.exe.)

  If it were to install on top of an existing gcc install, the result would be
"a nasty mess".  It's not the end of the world if someone who's willing to
play with an experimental compiler has to do a bit of rolling back and
forward, but I want to be very very very *very* certain that uploading won't
/under any circumstances/ trigger automatic installation on anyone who doesn't
specifically select it in the package chooser screen.


  If there's any danger of this happening, I will upload it as a "test"
version, and people will just have to deal with selecting all/a-valid-subset
of the packages themselves.  But if it's absolutely positively definitely
guaranteed safe, I'd rather upload it as the "curr" version, so the
interdependencies work without users having to click the "exp" radio button
and maybe get lots of other experimental stuff they don't really want.


    cheers,
      DaveK
-- 
Can't think of a witty .sigline today....

Re: New package behaviour of setup.ini

[Christopher Faylor]

On Tue, Sep 09, 2008 at 06:45:51PM +0100, Dave Korn wrote:
>
>
>  I want to make /very/ sure of this, which is why I'm asking:
>
>  If an entirely new package appears on the mirrors, setup.exe won't install
>it automatically, even if in the past you have maybe set the entire "Devel"
>category (as it happens to be in this case, but I could have said the entire
>top-level "All" category) to "Install"?  Or will it?

Selecting the Devel category is only a one-time thing AFAIK.  I don't
think that selection is stored in setup's crude disk settings files.

>  If there's any danger of this happening, I will upload it as a "test"
>version, and people will just have to deal with selecting all/a-valid-subset
>of the packages themselves.  But if it's absolutely positively definitely
>guaranteed safe, I'd rather upload it as the "curr" version, so the
>interdependencies work without users having to click the "exp" radio button
>and maybe get lots of other experimental stuff they don't really want.

I'm pretty sure it should be safe.  If it isn't then I'm sure the Cygwin users
will be intell...  Wait a minute.  Er.

I'm pretty sure it should be safe.  If it isn't then I'll be standing by
to make an emergency upset run to fix the problem.

Btw, does this version remove the -mno-cygwin option?  If so, I may have
to leave work early so that I can have a celebratory beverage.

cgf

RE: New package behaviour of setup.ini

[Dave Korn]

Christopher Faylor wrote on 09 September 2008 18:51:

> On Tue, Sep 09, 2008 at 06:45:51PM +0100, Dave Korn wrote:

> Btw, does this version remove the -mno-cygwin option?  If so, I may have
> to leave work early so that I can have a celebratory beverage.

  Sorry :( no beer for you today.  It's a fairly straight build of released
4.3.0 with just a bit of patching for shared libgcc etc.

  Anyway, I'll be doing the upload later tonight after I get home from work.
I thought it unlikely that the category settings would be persisted, but don't
want to get caught out by making a dumb assumption.... if at all possible ... 

 ;-)

    cheers,
      DaveK
-- 
Can't think of a witty .sigline today....

Re: New package behaviour of setup.ini

[Christopher Faylor]

On Tue, Sep 09, 2008 at 07:01:31PM +0100, Dave Korn wrote:
>Christopher Faylor wrote on 09 September 2008 18:51:
>>On Tue, Sep 09, 2008 at 06:45:51PM +0100, Dave Korn wrote:
>>Btw, does this version remove the -mno-cygwin option?  If so, I may
>>have to leave work early so that I can have a celebratory beverage.
>
>Sorry :( no beer for you today.  It's a fairly straight build of
>released 4.3.0 with just a bit of patching for shared libgcc etc.
>
>Anyway, I'll be doing the upload later tonight after I get home from
>work.  I thought it unlikely that the category settings would be
>persisted, but don't want to get caught out by making a dumb
>assumption....  if at all possible ...

I thought the next version of gcc was not going to perpetuate the
-mno-cygwin madness.  I'd really prefer not releasing a version with
this option even if it is just commented out.

cgf

Re: New package behaviour of setup.ini

[Christopher Faylor]

On Tue, Sep 09, 2008 at 02:35:23PM -0400, Christopher Faylor wrote:
>On Tue, Sep 09, 2008 at 07:01:31PM +0100, Dave Korn wrote:
>>Christopher Faylor wrote on 09 September 2008 18:51:
>>>On Tue, Sep 09, 2008 at 06:45:51PM +0100, Dave Korn wrote:
>>>Btw, does this version remove the -mno-cygwin option?  If so, I may
>>>have to leave work early so that I can have a celebratory beverage.
>>
>>Sorry :( no beer for you today.  It's a fairly straight build of
>>released 4.3.0 with just a bit of patching for shared libgcc etc.
>>
>>Anyway, I'll be doing the upload later tonight after I get home from
>>work.  I thought it unlikely that the category settings would be
>>persisted, but don't want to get caught out by making a dumb
>>assumption....  if at all possible ...
>
>I thought the next version of gcc was not going to perpetuate the
>-mno-cygwin madness.  I'd really prefer not releasing a version with
>this option even if it is just commented out.

Btw, Dave, do you want me to sponsor you as one of the new
cygwin-on-windows maintainers?  I was suggesting Kai Tietz as another
alternative for MinGW.  It would be nice to have an active Cygwin
maintainer as well.

cgf

RE: New package behaviour of setup.ini

[Dave Korn]

Christopher Faylor wrote on 09 September 2008 19:59:

> On Tue, Sep 09, 2008 at 02:35:23PM -0400, Christopher Faylor wrote:
>> On Tue, Sep 09, 2008 at 07:01:31PM +0100, Dave Korn wrote:
>>> Christopher Faylor wrote on 09 September 2008 18:51:
>>>> On Tue, Sep 09, 2008 at 06:45:51PM +0100, Dave Korn wrote:
>>>> Btw, does this version remove the -mno-cygwin option?  If so, I may
>>>> have to leave work early so that I can have a celebratory beverage.
>>> 
>>> Sorry :( no beer for you today.  It's a fairly straight build of
>>> released 4.3.0 with just a bit of patching for shared libgcc etc.
>>> 
>>> Anyway, I'll be doing the upload later tonight after I get home from
>>> work.  I thought it unlikely that the category settings would be
>>> persisted, but don't want to get caught out by making a dumb
>>> assumption....  if at all possible ...
>> 
>> I thought the next version of gcc was not going to perpetuate the
>> -mno-cygwin madness.  I'd really prefer not releasing a version with
>> this option even if it is just commented out.

  This is truely experimental stuff, but if it works well I'm planning to move
very fast to a 4.3.2 release, so I'll patch it out there; but I've kept
everyone waiting more than long enough and have limited free time.

> Btw, Dave, do you want me to sponsor you as one of the new
> cygwin-on-windows maintainers?  I was suggesting Kai Tietz as another
> alternative for MinGW.  It would be nice to have an active Cygwin
> maintainer as well.

  Yes please!  Modulo the above mentioned resource constraints, but I'm still
more than glad to help out.

  Stand back.  Here it comes ...

    cheers,
      DaveK
-- 
Can't think of a witty .sigline today....

Re: New package behaviour of setup.ini

[Christopher Faylor]

On Wed, Sep 10, 2008 at 10:00:53AM +0100, Dave Korn wrote:
>> On Tue, Sep 09, 2008 at 02:35:23PM -0400, Christopher Faylor wrote:
>>> I thought the next version of gcc was not going to perpetuate the
>>> -mno-cygwin madness.  I'd really prefer not releasing a version with
>>> this option even if it is just commented out.
>
>This is truely experimental stuff, but if it works well I'm planning to
>move very fast to a 4.3.2 release, so I'll patch it out there; but I've
>kept everyone waiting more than long enough and have limited free time.

I don't mean to sound ungrateful but I thought a big motivator for this
release was to remove -mno-cygwin.  Having YA release with it activated
is bound to generate confusion.

cgf

Re: New package behaviour of setup.ini

[Christopher Faylor]

On Wed, Sep 10, 2008 at 09:57:54AM -0400, Christopher Faylor wrote:
>On Wed, Sep 10, 2008 at 10:00:53AM +0100, Dave Korn wrote:
>>> On Tue, Sep 09, 2008 at 02:35:23PM -0400, Christopher Faylor wrote:
>>>> I thought the next version of gcc was not going to perpetuate the
>>>> -mno-cygwin madness.  I'd really prefer not releasing a version with
>>>> this option even if it is just commented out.
>>
>>This is truely experimental stuff, but if it works well I'm planning to
>>move very fast to a 4.3.2 release, so I'll patch it out there; but I've
>>kept everyone waiting more than long enough and have limited free time.
>
>I don't mean to sound ungrateful

...but I did anyway.  Sorry.  Whether there's a -mno-cygwin in this
thing or not this is a tremendous accomplishment.

Thank you for making this package available.  You are going to make a lot
of people happy.

cgf

RE: New package behaviour of setup.ini

[Dave Korn]

Christopher Faylor wrote on 10 September 2008 15:06:

> On Wed, Sep 10, 2008 at 09:57:54AM -0400, Christopher Faylor wrote:
>> On Wed, Sep 10, 2008 at 10:00:53AM +0100, Dave Korn wrote:
>>>> On Tue, Sep 09, 2008 at 02:35:23PM -0400, Christopher Faylor wrote:
>>>>> I thought the next version of gcc was not going to perpetuate the
>>>>> -mno-cygwin madness.  I'd really prefer not releasing a version with
>>>>> this option even if it is just commented out.
>>> 
>>> This is truely experimental stuff, but if it works well I'm planning to
>>> move very fast to a 4.3.2 release, so I'll patch it out there; but I've
>>> kept everyone waiting more than long enough and have limited free time.
>> 
>> I don't mean to sound ungrateful
> 
> ...but I did anyway.  Sorry.  Whether there's a -mno-cygwin in this
> thing or not this is a tremendous accomplishment.

  Nah, that's ok.

  On the plus side, at least if people try it, it won't work ;-)


    cheers,
      DaveK
-- 
Can't think of a witty .sigline today....

RE: New package behaviour of setup.ini

[Dave Korn]

Christopher Faylor wrote on 10 September 2008 14:58:

> I don't mean to sound ungrateful but I thought a big motivator for this
> release was to remove -mno-cygwin.  Having YA release with it activated
> is bound to generate confusion.

  You will be pleasantly surprised when I update to 4.3.2 sometime tonight or
tomorrow.  :)


    cheers,
      DaveK
-- 
Can't think of a witty .sigline today....

Re: New package behaviour of setup.ini

[Christopher Faylor]

On Thu, Sep 11, 2008 at 10:27:16AM +0100, Dave Korn wrote:
>Christopher Faylor wrote on 10 September 2008 14:58:
>>I don't mean to sound ungrateful but I thought a big motivator for this
>>release was to remove -mno-cygwin.  Having YA release with it activated
>>is bound to generate confusion.
>
>You will be pleasantly surprised when I update to 4.3.2 sometime
>tonight or tomorrow.  :)

The bottle is chilling right now.

cgf

Re: New package behaviour of setup.ini

[Albrecht Schlosser]

Dave Korn wrote:
> Christopher Faylor wrote on 10 September 2008 14:58:
> 
>> I don't mean to sound ungrateful but I thought a big motivator for this
>> release was to remove -mno-cygwin.  Having YA release with it activated
>> is bound to generate confusion.
> 
>   You will be pleasantly surprised when I update to 4.3.2 sometime tonight or
> tomorrow.  :)

But what will be the alternative for those who are using -mno-cygwin?

Will there be a simple (cross compile) replacement, as discussed previously.

I'm really curious (and concerned) about this, because I _am_ using this 
great cygwin toolkit on a daily basis, but I do also create non-cywgin
(native windows) programs. I wouldn't like the idea to have to install 
MinGW, too.

Thanks for your efforts and your hard work.

Albrecht

RE: New package behaviour of setup.ini

[Dave Korn]

Albrecht Schlosser wrote on 12 September 2008 08:51:

> Dave Korn wrote:
>> Christopher Faylor wrote on 10 September 2008 14:58:
>> 
>>> I don't mean to sound ungrateful but I thought a big motivator for this
>>> release was to remove -mno-cygwin.  Having YA release with it activated
>>> is bound to generate confusion.
>> 
>>   You will be pleasantly surprised when I update to 4.3.2 sometime
>> tonight or tomorrow.  :)
> 
> But what will be the alternative for those who are using -mno-cygwin?
> 
> Will there be a simple (cross compile) replacement, as discussed
> previously. 

  Yes!  Write "i586-pc-mingw32-gcc" instead of "gcc -mno-cygwin" and you're
done.  End of story; it really will be that straightforward.

    cheers,
      DaveK
-- 
Can't think of a witty .sigline today....

Re: New package behaviour of setup.ini

[Ralph Hempel]

Dave Korn wrote:

>> Will there be a simple (cross compile) replacement, as discussed
>> previously. 
> 
>   Yes!  Write "i586-pc-mingw32-gcc" instead of "gcc -mno-cygwin" and you're
> done.  End of story; it really will be that straightforward.

I've turned this text into a .BMP file and made it my desktop background
image :-)

Ralph