This should really fix #25675.
Please test, TIA!
https://github.com/wxWidgets/wxWidgets/pull/26200
(4 files)
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you are subscribed to this thread.![]()
@vadz pushed 1 commit.
—
View it on GitHub or unsubscribe.
You are receiving this because you are subscribed to this thread.![]()
This works great for my use case (the NASA GMAT project). After doing a make install, the dylibs in the install directory have the correct install names.
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you are subscribed to this thread.![]()
Thanks for testing!
I'm going to push this soon if there are no other comments.
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you are subscribed to this thread.![]()
@vads @ravidavi I haven't looked closely at the logic, but this patch seems to work for the wxwidgets 3.2.9 conda-forge package - thanks!
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you are subscribed to this thread.![]()
Thanks for confirmation, I'll push this to master too soon, so that it becomes part of the upcoming 3.3.2.
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you are subscribed to this thread.![]()
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you are subscribed to this thread.![]()
Hmm, I'm getting
/Users/xcodeserver/actions-runner/_work/wxWidgets/wxWidgets/change-install-names: line 11: realpath: command not found
in this CI run under macOS 12.6.9 so it's clearly not available in this version and we need make more changes to avoid breaking building there :-(
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you are subscribed to this thread.![]()
It sounds like readlink -f was added in 12.3, and realpath was added in 13.0: https://apple.stackexchange.com/questions/450035/is-the-unix-realpath-command-distributed-with-macos-ventura#comment689104_450116
Sorry I didn't check (although in my defence, Apple does not make things easy :/).
How about going back to basics? cd+dirname+pwd? https://github.com/JeffersonLab/coatjava/pull/120/changes
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you are subscribed to this thread.![]()
Using $(readlink ${build_libdir}/$(readlink ${build_libdir}/${libname})) should work, I'll push this to master soon, please retest before I backport this to 3.2.
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you are subscribed to this thread.![]()
Using $(readlink ${build_libdir}/$(readlink ${build_libdir}/${libname})) should work, I'll push this to master soon, please retest before I backport this to 3.2.
Won't that throw an error if ${build_libdir} is not a symbolic link?
➜ ~ pwd
/Users/paulmc
➜ ~ readlink /Users/paulmc/; echo $?
1
From the 15.5 manpage:
When invoked as readlink, only the target of the symbolic link is printed. If the given argument is not a symbolic link and the -f option is not specified, readlink will print nothing and exit with an error.
I don't have access to older versions, but https://ss64.com/mac/stat.html says:
When invoked as readlink, only the target of the symbolic link is printed. If the given argument is not a symbolic link, readlink will print nothing and exit with an error.
What a pain!
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you are subscribed to this thread.![]()
Using
$(readlink ${build_libdir}/$(readlink ${build_libdir}/${libname}))should work, I'll push this to master soon, please retest before I backport this to 3.2.Won't that throw an error if
${build_libdir}is not a symbolic link?
All path components don't have to be symlinks, only the last one does, and this is the case here, so in theory this should work. But more testing would be very welcome (I've pushed my changes now).
What a pain!
I can only agree with this.
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you are subscribed to this thread.![]()
How about adding this at the top?
resolve_path() {
# macos >= 13.0
if which realpath > /dev/null; then
cmd="realpath"
# macos < 13.0
else
cmd="readlink"
fi
$cmd $1
}
resolve_path $1
I've opened #26230, and am presently testing on macOS 15.5
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you are subscribed to this thread.![]()
All path components don't have to be symlinks, only the last one does, and this is the case here, so in theory this should work. But more testing would be very welcome (I've pushed my changes now).
Aah, no worries - I've closed #26320, and will test your readlink (readlink) suggestion.
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you are subscribed to this thread.![]()
No, this won't work at all, the commands don't return the same result, one returns the full path while another just returns the link target. I wish it were as samples as this...
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you are subscribed to this thread.![]()