Hi everyone, The latest git-core Portfile makes the port rcs a runtime dependency to git-core but there is no such port called rcs, at least port search rcs doesn't show up. Do you smell a broken port? Yuen Ho Wong
On Jan 25, 2007, at 23:03, Yuen Ho Wong wrote:
The latest git-core Portfile makes the port rcs a runtime dependency to git-core but there is no such port called rcs, at least port search rcs doesn't show up. Do you smell a broken port?
I recently committed the update from git-core 1.4.4.1 to 1.4.4.4. My comments: - The dependency on rcs has already existed in the portfile since the git 0.99.9b update was committed on 2005-11-05 in revision 14969; I did not just add it. - The git-core port built just fine for me. - The git-core port does not depend on a *port* rcs, but on there existing a *binary* named rcs. On my system, rcs exists as /usr/bin/rcs Are you experiencing difficulties building git-core? If so, please show us the error messages you get, and see whether you have a /usr/ bin/rcs on your system.
On Jan 25, 2007, at 10:03 PM, Yuen Ho Wong wrote:
Hi everyone,
The latest git-core Portfile makes the port rcs a runtime dependency to git-core but there is no such port called rcs, at least port search rcs doesn't show up. Do you smell a broken port?
Yeah, that's been broken since at least April 2006: <http://opendarwin.org/pipermail/darwinports/2006-April/033325.html> Bryan
Yuen Ho Wong
On Jan 26, 2007, at 01:26, Bryan Blackburn wrote:
The latest git-core Portfile makes the port rcs a runtime dependency to git-core but there is no such port called rcs, at least port search rcs doesn't show up. Do you smell a broken port?
Yeah, that's been broken since at least April 2006:
<http://opendarwin.org/pipermail/darwinports/2006-April/033325.html>
But what exactly is broken? As I said, git-core compiles just fine for me, since a binary called rcs, as requested by the port, does already exist on my system. Does it not exist on yours? What error do you see when you try to compile git-core?
On Jan 26, 2007, at 12:59 AM, Ryan Schmidt wrote:
On Jan 26, 2007, at 01:26, Bryan Blackburn wrote:
The latest git-core Portfile makes the port rcs a runtime dependency to git-core but there is no such port called rcs, at least port search rcs doesn't show up. Do you smell a broken port?
Yeah, that's been broken since at least April 2006:
<http://opendarwin.org/pipermail/darwinports/2006-April/033325.html>
But what exactly is broken? As I said, git-core compiles just fine for me, since a binary called rcs, as requested by the port, does already exist on my system. Does it not exist on yours? What error do you see when you try to compile git-core?
You're right, it should compile just fine with a normal 10.4 install (I'm not sure if rcs, in the BSD package on 10.4, was also in BSD in 10.3, which I think was still optional). The only issue is that the dependency claims that it can be resolved by installing the rcs port, which doesn't exist. With 10.4, we can definitely assume it's just there, so the fix is simply to remove that dep altogether. Though, if you look at that thread, most of the issues didn't seem to be of interest... Bryan
On Jan 26, 2007, at 13:34, Bryan Blackburn wrote:
On Jan 26, 2007, at 12:59 AM, Ryan Schmidt wrote:
On Jan 26, 2007, at 01:26, Bryan Blackburn wrote:
The latest git-core Portfile makes the port rcs a runtime dependency to git-core but there is no such port called rcs, at least port search rcs doesn't show up. Do you smell a broken port?
Yeah, that's been broken since at least April 2006:
<http://opendarwin.org/pipermail/darwinports/2006-April/033325.html>
But what exactly is broken? As I said, git-core compiles just fine for me, since a binary called rcs, as requested by the port, does already exist on my system. Does it not exist on yours? What error do you see when you try to compile git-core?
You're right, it should compile just fine with a normal 10.4 install (I'm not sure if rcs, in the BSD package on 10.4, was also in BSD in 10.3, which I think was still optional). The only issue is that the dependency claims that it can be resolved by installing the rcs port, which doesn't exist.
Ah, is that what it means? The portfile says "depends_run ... bin:merge:rcs ..." Does that mean "I need the binary called merge, and if there isn't one, install the port rcs"? I'm still muddling my way through the portfile syntax.
With 10.4, we can definitely assume it's just there, so the fix is simply to remove that dep altogether. Though, if you look at that thread, most of the issues didn't seem to be of interest...
On Jan 26, 2007, at 1:42 PM, Ryan Schmidt wrote:
On Jan 26, 2007, at 13:34, Bryan Blackburn wrote:
...
You're right, it should compile just fine with a normal 10.4 install (I'm not sure if rcs, in the BSD package on 10.4, was also in BSD in 10.3, which I think was still optional). The only issue is that the dependency claims that it can be resolved by installing the rcs port, which doesn't exist.
Ah, is that what it means? The portfile says "depends_run ... bin:merge:rcs ..." Does that mean "I need the binary called merge, and if there isn't one, install the port rcs"? I'm still muddling my way through the portfile syntax.
Yup, that's it; this is the older dependency style, before port:portname became the preferred. In fact, had this port been upgraded to port:rcs, then it'd definitely break. I think at this point the few uses for the old bin:program:portname and lib:libname:portname is for big things which we accept as valid pre- installed dependencies like X11. Bryan
With 10.4, we can definitely assume it's just there, so the fix is simply to remove that dep altogether. Though, if you look at that thread, most of the issues didn't seem to be of interest...
Citando Bryan Blackburn :
On Jan 26, 2007, at 1:42 PM, Ryan Schmidt wrote:
On Jan 26, 2007, at 13:34, Bryan Blackburn wrote:
...
You're right, it should compile just fine with a normal 10.4 install (I'm not sure if rcs, in the BSD package on 10.4, was also in BSD in 10.3, which I think was still optional). The only issue is that the dependency claims that it can be resolved by installing the rcs port, which doesn't exist.
Ah, is that what it means? The portfile says "depends_run ... bin:merge:rcs ..." Does that mean "I need the binary called merge, and if there isn't one, install the port rcs"? I'm still muddling my way through the portfile syntax.
Yup, that's it; this is the older dependency style, before port:portname became the preferred. In fact, had this port been upgraded to port:rcs, then it'd definitely break. I think at this point the few uses for the old bin:program:portname and lib:libname:portname is for big things which we accept as valid pre- installed dependencies like X11.
In fact, this dependency scheme permits to require something that may be provided by different ports and a preference if it is not found. I don't think it is really used like that but specifying bin:psql:postgresql8 would probably be a good idea for dealing with the many different postgres versions that exist. I have used it in the signing-party port because there are 3 different gnupg ports which probably all install a gpg binary (which is sufficient ofr signing-party). Emmanuel
participants (4)
-
Bryan Blackburn
-
Emmanuel Hainry
-
Ryan Schmidt
-
Yuen Ho Wong