Re: gEDA-user: Icarus Verilog going git.

[Dave McGuire <mcguire@xxxxxxxxxxxxx>]

On Jun 16, 2007, at 1:19 AM, Samuel A. Falvo II wrote:
>>    It was not compilation time.  I'm not running Linux on a PC, and
>> apparently the git authors think (or thought) that "all the world's a
>> Linux PC".
>
> Well, yeah, it was authored by Linus Torvalds, so that's to be
> expected.  I think it's gotten a _little_ looser since then, but it's
> still predominantly assuming a Posix-compatible environment.

   Not OS X or Solaris, at least not early last year.

>>    Yep, that's pretty close to what I ended up doing.  I seem to
>> recall, however, that it pulled down *all* revisions to every source
>> file.  (this was many months and a whole lot of stressful times ago,
>> so my memory of this is fuzzy)
>
> This *IS* true, and it does somewhat bother me too.  But I can see the
> rationale for it.
>
> To get around this, you have to tell git clone to grab a specific
> branch by name, kind of like you do with CVS.  Unlike CVS, however,
> git clone will grab _all_ branches by default, instead of just the
> master branch.

   Peter mentioned this in his response as well.  I can definitely  
see the advantage.  At the time, however, I just needed to *@!())#$%  
build FreeType (at least I think it was FreeType) and it was getting  
in my way.

>>    I've not seen darcs; I will take a look at it.  Admittedly CVS
>
> Darcs has a few bugs that prevents it from being useful for
> large-scale projects (but it's great for personal and small-scale
> stuff, which is how I use it).  Mercurial is something I've yet to
> check out, but apparently was strongly inspired by darcs.

   I sniffed at Mercurial a bit awhile back, at the suggestion of a  
friend...It seems to have potential.  If it was inspired by darcs, I  
really need to check out darcs.

>> need of "fixing".  CVS has its [rather significant] warts, no doubt
>> about it, but it does the job reasonably well.
>
> Well, SVN is designed to resolve those warts, and I've heard a lot of
> good things about it in this area.

   Me too...SVN does some things differently enough from CVS, though,  
that some relatively major things need to be "un-learned".  In SVN,  
for example, individual files don't have revision numbers...the  
entire repository does.  When you commit one file, THE revision  
number *for the entire repository* is incremented.  So basically, a  
CVS person (like me) would see a whole slew of "different version  
numbers of file foo.c" (think CVS here) that are...identical.

   Not "bad" per se, just very different from CVS.  And like it or  
not, CVS is the one with the huge installed base, that all those  
other packages are trying to replace.

> I have to admit, though, I'm hooked on git, darcs, and similar tools.
> Distributed SCMs do make _very_ rapid commits (I often commit at least
> 6 times an hour or more) trivially easy, and often have superior
> merging algorithms.  Hence, distributed version control systems really
> do facilitate a more agile way of developing software.

   I'm a frequent committer too...I find that smaller steps make it  
easier to backtrack if I need to, and (perhaps more importantly) make  
it easier to understand the "flow" of changes as code evolves.

> At work, we use Perforce, a SVN-like commercial system.  It's OK -- it
> works rather well as far as it goes.  But there is no way I could ever
> pull off 6+ commits per hour without utterly pissing off other
> developers in the process.  :)

   Oh My. :)  Is it a performance issue?

           -Dave

-- 
Dave McGuire
Port Charlotte, FL




_______________________________________________
geda-user mailing list
geda-user@xxxxxxxxxxxxxx
http://www.seul.org/cgi-bin/mailman/listinfo/geda-user