Bug#361740: liblua50-dev: symlink pointing to .
On Sun, 2006-04-09 at 18:23 -0700, Russ Allbery wrote: There is exactly one header file, and IMO packages should care enough about their dependencies to not do silly things to get a single file included. It smells like a transitional measure, but I don't know for sure. I see that it was intentional: It was indeed intentional I do generally agree that it's not the best idea if there are other solutions to the same problem, but I can see why one might arrive at this solution. And the exact reasoning is thus... Some programs are written aware of the lua50 and lua40 dirs used in Debian and other distros. They do: #include lua50/lua.h Some are aware and do: #include lua.h and have -I/usr/include/lua50 Some are aware and do: #include lua/lua.h and have -I/usr/include/lua50 Some are unaware and do either of the second without the -I -- at that point it's up to the packager to fix up the build either by adding a -I or by fixing the #include statements to be more correct. There is a wealth of software out there which uses Lua, and not all of it is suited to a packaged version of Lua and those which are are not always suited to the idea that there might be differing versions on the filesystem at the same time. This is a compatibility measure in an attempt to increase the amount of software which can build with as little intervention as possible from the packager. Also, it's not just one include file. It's just that others are in liblualib50-dev I imagine. Regards, Daniel -- Daniel Silverstonehttp://www.debian.org/ PGP mail accepted and encouraged.Key Id: 2BC8 4016 2068 7895 signature.asc Description: This is a digitally signed message part
Bug#361740: liblua50-dev: symlink pointing to .
Daniel Silverstone wrote: Some are aware and do: #include lua.h and have -I/usr/include/lua50 If software using Lua's C API wants to be portable, it should use the lua.h form and get the include path from pkg-config. That should be the end of it. Should someone want to use such software on a tiny system without pkg-config, they can make the one line change to the makefile. Hacks such as the symlink only encourage further obliviousness to best practice. --John -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#361740: liblua50-dev: symlink pointing to .
Package: liblua50-dev Version: 5.0.2-6 Severity: normal $ dpkg -c /var/cache/apt/archives/liblua50-dev_5.0.2-6_i386.deb |grep -- '-' lrwxrwxrwx root/root 0 2006-04-09 10:03:01 ./usr/lib/liblua50.so - liblua50.so.5.0 lrwxrwxrwx root/root 0 2006-04-09 10:03:01 ./usr/include/lua50/lua - . -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#361740: liblua50-dev: symlink pointing to .
Justin Pryzby [EMAIL PROTECTED] writes: Package: liblua50-dev Version: 5.0.2-6 Severity: normal $ dpkg -c /var/cache/apt/archives/liblua50-dev_5.0.2-6_i386.deb |grep -- '-' lrwxrwxrwx root/root 0 2006-04-09 10:03:01 ./usr/lib/liblua50.so - liblua50.so.5.0 lrwxrwxrwx root/root 0 2006-04-09 10:03:01 ./usr/include/lua50/lua - . Why is this a bug? I'm missing something, I think. -- Russ Allbery ([EMAIL PROTECTED]) http://www.eyrie.org/~eagle/ -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#361740: liblua50-dev: symlink pointing to .
On Sun, Apr 09, 2006 at 04:30:06PM -0700, Russ Allbery wrote: Justin Pryzby [EMAIL PROTECTED] writes: Package: liblua50-dev Version: 5.0.2-6 Severity: normal $ dpkg -c /var/cache/apt/archives/liblua50-dev_5.0.2-6_i386.deb |grep -- '-' lrwxrwxrwx root/root 0 2006-04-09 10:03:01 ./usr/lib/liblua50.so - liblua50.so.5.0 lrwxrwxrwx root/root 0 2006-04-09 10:03:01 ./usr/include/lua50/lua - . Why is this a bug? I'm missing something, I think. Is it deliberate?! Why would one need such a symlink? It bothered me because my grep -r /usr/include/ gave me a recusion warning.. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#361740: liblua50-dev: symlink pointing to .
Justin Pryzby [EMAIL PROTECTED] writes: On Sun, Apr 09, 2006 at 04:30:06PM -0700, Russ Allbery wrote: Justin Pryzby [EMAIL PROTECTED] writes: Package: liblua50-dev Version: 5.0.2-6 Severity: normal $ dpkg -c /var/cache/apt/archives/liblua50-dev_5.0.2-6_i386.deb |grep -- '-' lrwxrwxrwx root/root 0 2006-04-09 10:03:01 ./usr/lib/liblua50.so - liblua50.so.5.0 lrwxrwxrwx root/root 0 2006-04-09 10:03:01 ./usr/include/lua50/lua - . Why is this a bug? I'm missing something, I think. Is it deliberate?! Why would one need such a symlink? It bothered me because my grep -r /usr/include/ gave me a recusion warning.. Does grep -r follow symlinks? That sounds kind of dangerous. I haven't ever used or uploaded the package, so I'm not sure, but my guess is that it's so that software built with -I/usr/include/lua50 that includes files like #include lua/foo.h will find the header files. There are other ways to accomplish the same thing, but I would have been tempted to do that myself, if that is indeed the problem being solved. -- Russ Allbery ([EMAIL PROTECTED]) http://www.eyrie.org/~eagle/ -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#361740: liblua50-dev: symlink pointing to .
On Sun, Apr 09, 2006 at 05:54:03PM -0700, Russ Allbery wrote: Justin Pryzby [EMAIL PROTECTED] writes: On Sun, Apr 09, 2006 at 04:30:06PM -0700, Russ Allbery wrote: Justin Pryzby [EMAIL PROTECTED] writes: Package: liblua50-dev Version: 5.0.2-6 Severity: normal $ dpkg -c /var/cache/apt/archives/liblua50-dev_5.0.2-6_i386.deb |grep -- '-' lrwxrwxrwx root/root 0 2006-04-09 10:03:01 ./usr/lib/liblua50.so - liblua50.so.5.0 lrwxrwxrwx root/root 0 2006-04-09 10:03:01 ./usr/include/lua50/lua - . Why is this a bug? I'm missing something, I think. Is it deliberate?! Why would one need such a symlink? It bothered me because my grep -r /usr/include/ gave me a recusion warning.. Does grep -r follow symlinks? That sounds kind of dangerous. It does, I just checked. I'm not sure why it's dangerous, xargs in /tmp/ is dumb, and anywhere else you control and should be safe.. I haven't ever used or uploaded the package, so I'm not sure, but my guess is that it's so that software built with -I/usr/include/lua50 that includes files like #include lua/foo.h will find the header files. There are other ways to accomplish the same thing, but I would have been tempted to do that myself, if that is indeed the problem being solved. There is exactly one header file, and IMO packages should care enough about their dependencies to not do silly things to get a single file included. It is in the .diff.gz, so apparently provided upstream. + (cd debian/liblua50-dev/usr/include/lua50; \ + ln -s . lua ) Daniel, can you comment on it? Justin -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#361740: liblua50-dev: symlink pointing to .
Justin Pryzby [EMAIL PROTECTED] writes: On Sun, Apr 09, 2006 at 05:54:03PM -0700, Russ Allbery wrote: Does grep -r follow symlinks? That sounds kind of dangerous. It does, I just checked. I'm not sure why it's dangerous, xargs in /tmp/ is dumb, and anywhere else you control and should be safe.. Well, because of exactly this. Symlinks that create circular directory structures are extremely common. Not dangerous in the security vulnerability sense, dangerous in the this may not do what you expect and be very slow and annoying while not doing it sense. There is exactly one header file, and IMO packages should care enough about their dependencies to not do silly things to get a single file included. It smells like a transitional measure, but I don't know for sure. I see that it was intentional: lua50 (5.0.2-3) unstable; urgency=low * Fold in a patch from Reuben Thomas, integrating the signal and error code from Fedora Core. Thanks Reuben. * Lintian cleanups (recursive symlink stays, sorry) -- Daniel Silverstone [EMAIL PROTECTED] Thu, 3 Jun 2004 14:35:00 -0300 I do generally agree that it's not the best idea if there are other solutions to the same problem, but I can see why one might arrive at this solution. -- Russ Allbery ([EMAIL PROTECTED]) http://www.eyrie.org/~eagle/ -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#361740: liblua50-dev: symlink pointing to .
On Sun, Apr 09, 2006 at 06:23:00PM -0700, Russ Allbery wrote: Justin Pryzby [EMAIL PROTECTED] writes: On Sun, Apr 09, 2006 at 05:54:03PM -0700, Russ Allbery wrote: Does grep -r follow symlinks? That sounds kind of dangerous. It does, I just checked. I'm not sure why it's dangerous, xargs in /tmp/ is dumb, and anywhere else you control and should be safe.. Well, because of exactly this. Symlinks that create circular directory structures are extremely common. Not dangerous in the security vulnerability sense, dangerous in the this may not do what you expect and be very slow and annoying while not doing it sense. There is exactly one header file, and IMO packages should care enough about their dependencies to not do silly things to get a single file included. It smells like a transitional measure, but I don't know for sure. I see that it was intentional: lua50 (5.0.2-3) unstable; urgency=low * Fold in a patch from Reuben Thomas, integrating the signal and error code from Fedora Core. Thanks Reuben. * Lintian cleanups (recursive symlink stays, sorry) -- Daniel Silverstone [EMAIL PROTECTED] Thu, 3 Jun 2004 14:35:00 -0300 Damn, I looked for this and bet I saw it too, but ignored it because it m/^Lintian/.. I can see that it wont go away easily, since there are ~50 depending packages, any of which might be including strangely named files.. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
