www.digitalmars.com         C & C++   DMDScript  

digitalmars.D - patched phobos and phobos_ext

reply UB <ugis.baumanis gmail.com> writes:
Is anybody interested in patched phobos and phobos_ext ( Tango's (?)
containers, ..., etc.), if Walter and Tango guys has nothing against it?
The main idea (from users' point of view) is download source code from dsource
(?),
and build locally phobos and phobos_ext.
With your help I am ready to spend some time on this.
Are here phobos users, or people are moving mostly toward Tango? Any opinions?

ubau
Oct 02 2007
next sibling parent reply "Vladimir Panteleev" <thecybershadow gmail.com> writes:
On Wed, 03 Oct 2007 00:23:53 +0300, UB <ugis.baumanis gmail.com> wrote:


 Are here phobos users, or people are moving mostly toward Tango? Any opinions?


I tend to use Phobos unless I need a feature of Tango. I still have a few pet
peeves with Tango (mainly the neverending import clauses, the "you must create
an object for everything" ideology and generally the involved verbosity).

-- 
Best regards,
 Vladimir                          mailto:thecybershadow gmail.com
Oct 02 2007
parent reply UB <ugis.baumanis gmail.com> writes:
== Quote from Vladimir Panteleev (thecybershadow gmail.com)'s article

 On Wed, 03 Oct 2007 00:23:53 +0300, UB <ugis.baumanis gmail.com> wrote:

 Are here phobos users, or people are moving mostly toward Tango? Any opinions?


 I tend to use Phobos unless I need a feature of Tango. I still have a few pet


peeves with Tango (mainly the neverending import clauses, the "you must create
an
object for everything" ideology and generally the involved verbosity).

What are these features used of Tango?
Are they deep related to Tango, or could be used separately with Phobos?

ubau
Oct 02 2007
parent "Vladimir Panteleev" <thecybershadow gmail.com> writes:
On Wed, 03 Oct 2007 01:30:38 +0300, UB <ugis.baumanis gmail.com> wrote:


 =3D=3D Quote from Vladimir Panteleev (thecybershadow gmail.com)'s arti=


cle

 On Wed, 03 Oct 2007 00:23:53 +0300, UB <ugis.baumanis gmail.com> wrot=




e:

 Are here phobos users, or people are moving mostly toward Tango? An=






y opinions?

 I tend to use Phobos unless I need a feature of Tango. I still have a=




 few pet

 peeves with Tango (mainly the neverending import clauses, the "you mus=


t create an

 object for everything" ideology and generally the involved verbosity).=



 What are these features used of Tango?
 Are they deep related to Tango, or could be used separately with Phobo=


s?

I use the net package. I would love to see something like =

string httpGet(string url, string[string] extraHeaders =3D null);
string httpPost(string url, void[] data, string[string] extraHeaders =3D=
 null);

(and, possibly, functions/classes that these functions wrap around with =
more control)

Another thing was that it had more complete coverage of the Win32 API, b=
ut since I discovered the WindowsAPI[1] bindings package, that's hardly =
an issue.

I think that's about it. One thing I do love about Tango, though, is tha=
t they're much faster at fixing bugs than Phobos :D

[1] http://www.dsource.org/projects/bindings/wiki/WindowsApi

-- =

Best regards,
 Vladimir                          mailto:thecybershadow gmail.com
Oct 02 2007
prev sibling parent reply "David Wilson" <dw botanicus.net> writes:
On 02/10/2007, UB <ugis.baumanis gmail.com> wrote:


 Is anybody interested in patched phobos and phobos_ext ( Tango's (?)
 containers, ..., etc.), if Walter and Tango guys has nothing against it?


I would be greatly interested in a distribution of Phobos that
includes critical fixes (threading, GC, etc.) with minimal functional
enhancements. Someone recently mentioned having a definitive patch
list for Phobos; I think that at least is a great idea.

What I really believe D needs is a Phobos distribution ala. Net-Phobos
like Netqmail is for Qmail. If you're not familiar with that situation
- the Qmail author produced an excellent body of code, but has a
policy of keeping it incredibly traditional, such as not using
"errno.h", which causes the build to break on many modern C libraries
that virtualize the actual "errno" symbol.

Net-Qmail contains the minimal fixes required to restore Qmail
functionality in a modern environment, and as such has become the de
facto Qmail distribution. People trust it because they know it barely
deviates from the official source code.

Specifically I believe it would be a grand mistake to extend or
functionally modify Phobos in any way - if you expect any community
uptake as opposed to Yet Another Runtime.

GDC also made its own modifications to Phobos. It would be interesting
to enumerate them and try to provide a unified distribution - if
possible you might expect excellent uptake.



 The main idea (from users' point of view) is download source code from dsource
 (?),
 and build locally phobos and phobos_ext.


-1 for phobos_ext


 With your help I am ready to spend some time on this.


I think a patched distribution is a great idea and low-hanging fruit.
The patches already exist, all that needs done is creation of a
patched archive, documenting which patches were included, and a
reasonable amount of QA. (One bad release and people will begin to
lose trust).

Again I stress I believe it would be a grave mistake to extend Phobos.
There are enough "standard libraries" as it is. :)

Thanks,


David.
Oct 02 2007
parent reply "U.Baumanis" <ugis.baumanis gmail.com> writes:
On Wed, 03 Oct 2007 00:59:05 +0100, David Wilson wrote:


 On 02/10/2007, UB <ugis.baumanis gmail.com> wrote:
 

 Is anybody interested in patched phobos and phobos_ext ( Tango's (?)
 containers, ..., etc.), if Walter and Tango guys has nothing against it?


 
 I would be greatly interested in a distribution of Phobos that
 includes critical fixes (threading, GC, etc.) with minimal functional
 enhancements. Someone recently mentioned having a definitive patch
 list for Phobos; I think that at least is a great idea.
 
 What I really believe D needs is a Phobos distribution ala. Net-Phobos
 like Netqmail is for Qmail. If you're not familiar with that situation
 - the Qmail author produced an excellent body of code, but has a
 policy of keeping it incredibly traditional, such as not using
 "errno.h", which causes the build to break on many modern C libraries
 that virtualize the actual "errno" symbol.
 
 Net-Qmail contains the minimal fixes required to restore Qmail
 functionality in a modern environment, and as such has become the de
 facto Qmail distribution. People trust it because they know it barely
 deviates from the official source code.
 
 Specifically I believe it would be a grand mistake to extend or
 functionally modify Phobos in any way - if you expect any community
 uptake as opposed to Yet Another Runtime.
 
 GDC also made its own modifications to Phobos. It would be interesting
 to enumerate them and try to provide a unified distribution - if
 possible you might expect excellent uptake.
 
 
 
 Again I stress I believe it would be a grave mistake to extend Phobos.
 There are enough "standard libraries" as it is. :)
 
 Thanks,
 
 
 David.


Thanks for your ideas and advices.
Regarding phobos_ext, I agree - there is no need to mix two projects in
one. (this idea just came out of previous posts about relatively easy
extracting Tango's containers and using with Phobos)
So yes, main target must be Phobos and then other libs will hopefully use
it more.

And may be than Tango could be used as extension to Phobos.

ubau
Oct 02 2007
parent Brad Roberts <braddr puremagic.com> writes:
On Wed, 3 Oct 2007, U.Baumanis wrote:


 On Wed, 03 Oct 2007 00:59:05 +0100, David Wilson wrote:
 

 On 02/10/2007, UB <ugis.baumanis gmail.com> wrote:
 

 Is anybody interested in patched phobos and phobos_ext ( Tango's (?)
 containers, ..., etc.), if Walter and Tango guys has nothing against it?


 
 I would be greatly interested in a distribution of Phobos that
 includes critical fixes (threading, GC, etc.) with minimal functional
 enhancements. Someone recently mentioned having a definitive patch
 list for Phobos; I think that at least is a great idea.
 
 What I really believe D needs is a Phobos distribution ala. Net-Phobos
 like Netqmail is for Qmail. If you're not familiar with that situation
 - the Qmail author produced an excellent body of code, but has a
 policy of keeping it incredibly traditional, such as not using
 "errno.h", which causes the build to break on many modern C libraries
 that virtualize the actual "errno" symbol.
 
 Net-Qmail contains the minimal fixes required to restore Qmail
 functionality in a modern environment, and as such has become the de
 facto Qmail distribution. People trust it because they know it barely
 deviates from the official source code.
 
 Specifically I believe it would be a grand mistake to extend or
 functionally modify Phobos in any way - if you expect any community
 uptake as opposed to Yet Another Runtime.
 
 GDC also made its own modifications to Phobos. It would be interesting
 to enumerate them and try to provide a unified distribution - if
 possible you might expect excellent uptake.
 
 
 
 Again I stress I believe it would be a grave mistake to extend Phobos.
 There are enough "standard libraries" as it is. :)
 
 Thanks,
 
 
 David.


 
 Thanks for your ideas and advices.
 Regarding phobos_ext, I agree - there is no need to mix two projects in
 one. (this idea just came out of previous posts about relatively easy
 extracting Tango's containers and using with Phobos)
 So yes, main target must be Phobos and then other libs will hopefully use
 it more.
 
 And may be than Tango could be used as extension to Phobos.
 
 ubau
 


Let me put this thread to bed.

For each phobos bug that anyone cares to see fixed.  If you'll please make 
sure there's a bug filed in bugzilla and attach a fix to it with 
appropriate unit tests, I'll make sure it's folded into a dmd release.

This does _not_ include feature enhancements, just fixes to bugs.

Later,
Brad
Oct 02 2007