Opened 15 years ago
Closed 14 years ago
#24234 closed update (fixed)
Update cln from 1.2.2 to 1.3.1
| Reported by: | danmichaelo+macports@… | Owned by: | gwright@… |
|---|---|---|---|
| Priority: | Normal | Milestone: | |
| Component: | ports | Version: | |
| Keywords: | haspatch | Cc: | ryandesign (Ryan Carsten Schmidt) |
| Port: | cln |
Description
A request for an update to version 1.3.0 has been posted before (#20884). I provide a patch for updating to version 1.3.1.
I can't get universal build to work with either version 1.2.2 or 1.3.1, so I added "universal_variant no".
Attachments (2)
Change History (9)
Changed 15 years ago by danmichaelo+macports@…
| Attachment: | Portfile-cln.diff added |
|---|
comment:1 Changed 15 years ago by mf2k (Frank Schima)
| Keywords: | haspatch added |
|---|---|
| Owner: | changed from macports-tickets@… to gwright@… |
| Version: | 1.8.2 |
comment:2 follow-up: 3 Changed 15 years ago by ryandesign (Ryan Carsten Schmidt)
| Cc: | ryandesign@… added |
|---|
Changed 15 years ago by ryandesign (Ryan Carsten Schmidt)
reformatting of danmichaelo's patch
comment:3 Changed 15 years ago by ryandesign (Ryan Carsten Schmidt)
Replying to ryandesign@…:
"-O2" in the cflags apparently makes the test suite fail
I can't verify this; the cln 1.2.2 tests say they pass fine for me with -O2 on Snow Leopard x86_64, Leopard ppc, and Tiger i386. Perhaps there is another platform on which this still causes a failure.
Also, when updating cln to 1.3.x, all dependents of cln will need a revbump for the new library version number. At present, that seems to be GiNaC, libqalculate, and py26-clnum.
comment:4 follow-up: 5 Changed 15 years ago by danmichaelo+macports@…
Thanks for the update. I'm sorry for my non-informative description. Yes, I were able to install and run the tests suite with -O2. So my guess was that the need for the removal of -O2 had to do with a bug in an older version of cln or g++ (the CLN manual hints on some issues with -O2 and old versions of g++.. Also, the CLN manual warns about the -O3 flag, which was why I added the removal of it just in case. Since I'm quite new to MacPorts I do not know whether the defaults are overriden by the user's environment, etc. If you think -O2 always will be the default, then just remove the "-O2/-O3" part from the Portfile. My intention was just to make sure -O2 was chosen.
I filed a ticket on the update of Ginac as well: #24242. Unfortunately I did whitespace changes there as well... so I will upload a new version now.
comment:5 follow-up: 6 Changed 15 years ago by ryandesign (Ryan Carsten Schmidt)
Replying to danmichaelo+macports@…:
Yes, I were able to install and run the tests suite with -O2.
Compiled for what architecture, running on what version of Mac OS X?
So my guess was that the need for the removal of -O2 had to do with a bug in an older version of cln or g++
Shouldn't relate to an older version of cln, since the lines forcing the optimization level to -O were added to the portfile in r38679 at the same time as the version was updated to 1.2.2. Perhaps it was for an older version of g++, i.e. version 3.x which was on Panther. But MacPorts doesn't run on Panther anymore so if that's the case then we don't care about that anymore. Input from Greg would be appreciated here: where did you experience problems with the testsuite with -O2?
(the CLN manual hints on some issues with -O2 and old versions of g++..
But only on Sparc.
Also, the CLN manual warns about the -O3 flag, which was why I added the removal of it just in case. Since I'm quite new to MacPorts I do not know whether the defaults are overriden by the user's environment, etc. If you think -O2 always will be the default, then just remove the "-O2/-O3" part from the Portfile. My intention was just to make sure -O2 was chosen.
MacPorts clears the environment and sets up the defaults described in the guide. The default optimization level in MacPorts is -O2. See configure.optflags in the guide. I don't think there's a way for the user to change this default either at MacPorts compile time nor from macports.conf.
comment:6 Changed 15 years ago by danmichaelo+macports@…
Replying to ryandesign@…:
Replying to danmichaelo+macports@…:
Yes, I were able to install and run the tests suite with -O2.
Compiled for what architecture, running on what version of Mac OS X?
I'm able to build and test both 1.2.2 and 1.3.1 on Snow Leopard x86_64.
I'm neither able to build 1.2.2 nor 1.3.1 for i386 (universal), neither with -O nor -O2. But this problem is not with the tests, but the build failing, as described in #24674.
comment:7 Changed 14 years ago by ryandesign (Ryan Carsten Schmidt)
| Resolution: | → fixed |
|---|---|
| Status: | new → closed |
Greg updated the port to 1.3.1 in r71205. I don't know if he used the patches in this ticket.
Replying to danmichaelo+macports@…:
The patch is not acceptable as-is because it makes whitespace changes to the entire portfile. Please either supply diffs that make exclusively whitespace changes, or that make functional changes that respect the portfile's existing whitespace conventions. Attached is a patch that I believe represents the functional changes you're proposing while preserving the portfile's existing whitespace. Your changes amount to updating the version and checksums, changing the description and long description, and modifying the optimization flags. The latter concerns me because a) removing "-03" from the cflags doesn't make sense because "-O3" is not in the cflags by default; "-O2" is; and b) having "-O2" in the cflags apparently makes the test suite fail -- at least, that's why this was originally changed in r38679; I haven't yet tested to verify whether that was the case, or whether it continues to be the case now with 1.3.1. If you have already verified that this is no longer a problem, please let us know.
I filed #24674 for this issue.