[upforgrabs] tools/update

[Jacob Carlborg]

On 2011-06-10 22:09, Nick Sabalausky wrote:
> "Andrei Alexandrescu"<SeeWebsiteForEmail at erdani.org>  wrote in message
> news:ist957$19nj$1 at digitalmars.com...
>> We need a script to update all of dmd's paraphernalia. By this I'm putting
>> the task up for grabs and am describing it.
>>
>> The script would require git to be installed, and would orchestrate git
>> commands to freshen the installation. Either a tag can be given, or the
>> script would automatically detect the latest "semistable" release, or even
>> the script can be told to get the latest and greatest. The exact tagging
>> convention is to be defined later.
>>
>> The trick with such tools is they must "just work". That means wherever
>> and however you installed dmd, the script should be able to detect
>> everything it needs for a successful update. This means: dmd and rdmd
>> binary locations, druntime and phobos library locations, and druntime and
>> phobos import locations.
>>
>
> I've already been planning to add the features of "recompile DMD" and "check
> out DMD from git and compile it" to DVM (unless Jacob plans to beat me to
> it). You do have good suggestions on some of the implementation details. If
> there's still another feature in your proposal that I've overlooked here,
> let me know.

This has been on my long-term-todo list since I started the project. 
Although I have no idea if this is possible, because, as far as I know, 
there is no the "connection" between the DMD, druntime and Phobos 
repositories. How do I know if a particular revision/hash from one 
repository will work with one from the other repositories?

> I was going to wait to do this until Jacob merged in my Windows pull request
> (he said he'd get to it end of next week, IIRC), because I want to base it
> off a DVM that already has Windows support, but maybe I shouldn't wait and
> should just fork my "windows fork" and get to it? (I'm still fairly new to
> DVCS and "proper DVCS procedures".)
>
> DVM is written in D1/Tango, though. Would that be a problem? If so, and if
> Jacob doesn't have any objections to it being ported to D2/Phobos, then I'd
> volunteer to do the port. I already have experience porting a bunch of my
> own code from D1/Tango to D2/Phobos. But I don't want to be hasty about
> porting it: I assume Jacob had a reason for using D1/Tango. I don't know how
> important the reason was.

I would prefer to use Tango. The reasons I chose Tango were/are:

* I'm used to D1/Tango
* Every time I try to use D2 something isn't working
* I still think Tango is more superior than Phobos, although Phobos is 
doing a great job of filling in the feature holes
* A particular feature and the most important that I needed was to easy 
download files via HTTP, Tango makes this very easy. Now Phobos is 
getting a curl wrapper which should make it as easy, don't know the 
status of that though. I don't want to and don't know how to use a low 
level socket to do this.

-- 
/Jacob Carlborg