Re: Changes in handling library dependencies

2000-01-21 Thread Brian May
Brian == Brian May [EMAIL PROTECTED] writes: Jason You can link a library with missing symbols but when you Jason link a program to that library the linker will complain. Brian Yes, I see you are right. ldd shows libresolv (which was Brian included on the command line), but

Re: Changes in handling library dependencies

2000-01-21 Thread Jason Gunthorpe
On 21 Jan 2000, Brian May wrote: Hang on. You can't do it!!! At least, not with libtool. What? .la files are only for static linking, we are talking about dynamic. It is good that libtool complains : Look inside the .la file, it is just a text file. Jason

Re: Changes in handling library dependencies

2000-01-21 Thread Brian May
Jason == Jason Gunthorpe [EMAIL PROTECTED] writes: Jason On 21 Jan 2000, Brian May wrote: Hang on. You can't do it!!! At least, not with libtool. Jason What? .la files are only for static linking, we are talking Jason about dynamic. It is good that libtool complains :

Re: Changes in handling library dependencies

2000-01-21 Thread Jason Gunthorpe
On 21 Jan 2000, Brian May wrote: How would you do it? If you are using libtool and the .la has the correct information but the .so does not have the proper depends then I think we should start tormenting the libtool authors to fix it. It is not hard, you just throw the right -L and -l options

Re: Changes in handling library dependencies

2000-01-21 Thread Ben Collins
On Fri, Jan 21, 2000 at 12:41:03PM +1100, Brian May wrote: Jason == Jason Gunthorpe [EMAIL PROTECTED] writes: Jason On 21 Jan 2000, Brian May wrote: Hang on. You can't do it!!! At least, not with libtool. Jason What? .la files are only for static linking, we are talking

Re: Changes in handling library dependencies

2000-01-21 Thread Brian May
Mark == Mark Baker [EMAIL PROTECTED] writes: Mark On Thu, Jan 20, 2000 at 03:34:29PM +0100, Ronald van Loon Mark wrote: program needs on the command line. While it may be true that it is sufficient to be *dependent* only on imlib, it is still necessary to specify all those

Re: Custom undocumented(7)s are just as bad.

2000-01-21 Thread Seth R Arnold
On Thu, Jan 20, 2000 at 11:01:30PM -0300, Nicolás Lichtmaier wrote: I would expect that a mantainer have a little knowledge of his package. If a binary is not meant to be called by the user, it is a bug to have it in the PATH. Nicolas, this might show my naivite, but where should programs

Re: Changes in handling library dependencies

2000-01-21 Thread Brian May
Jason == Jason Gunthorpe [EMAIL PROTECTED] writes: Jason gcc -L ./libs -lkerb5 -o libgssapi.so [...] Like I said earlier, that doesn't work if you put libtool in front of the command line. -- Brian May [EMAIL PROTECTED]

Re: Changes in handling library dependencies

2000-01-21 Thread Jason Gunthorpe
On 21 Jan 2000, Brian May wrote: Jason gcc -L ./libs -lkerb5 -o libgssapi.so [...] Like I said earlier, that doesn't work if you put libtool in front of the command line. If libtool doesn't do that for you but writes that info to a .la file then it is plain and simple Broken(tm) for

Re: Changes in handling library dependencies

2000-01-21 Thread Brian May
Ben == Ben Collins [EMAIL PROTECTED] writes: Ben Shared library information in the .la is just redundancy of Ben what info can be gotten from the library itself. It is only Ben useful to the human reader (who understands little about ldd Ben and objdump). Ld does not use it, and

Re: Changes in handling library dependencies

2000-01-21 Thread Brian May
Brian == Brian May [EMAIL PROTECTED] writes: Brian From the documentation: Brian(1) However, you should never use `-L' or `-l' flags to Brian link against an uninstalled libtool library. Just specify Brian the relative path to the `.la' file, such as Brian

Ban undocumented(7) for native packages?

2000-01-21 Thread Chris Lawrence
I think undocumented(7) does serve a useful purpose, particularly when the program does have useful documentation in another format. However, I also think policy should prohibit the use of undocumented(7) by Debian native packages. Debian has made a statement (elsewhere in the policy documents)

RE: Changes in handling library dependencies

2000-01-21 Thread Ronald van Loon
No, this is wrong. With dynamic linking it is proper to specify ONLY the libraries whos functions you explicitly use. For instance if you only call imlib functions then you should link only to imlib. In turn, imlib's shared library can link to or dlopen whatever other shared libraries it

Re: Changes in handling library dependencies

2000-01-21 Thread Roman Hodek
Then explain to my why it worked perfectly for me even if I did not list all those libraries in the link command? Seems like the linker is smart enough.. Yes, it is. (But only for shared libs, of course, static libs have no NEEDED entries.) Roman

Re: Changes in handling library dependencies

2000-01-21 Thread Roman Hodek
*static* libraries are conisiderably different and in that case you do need to mention all of the libraries that all of the sub libraries touch. However, I thought gnome used libtool which takes care of that through its .la files, making this whole point moot. Yep, libtool tries to overcome

RE: Changes in handling library dependencies

2000-01-21 Thread Jason Gunthorpe
On Fri, 21 Jan 2000, Ronald van Loon wrote: had to request an address to load your library. Things have come quite a ways since then. But I digress. Welcome to ELF.. I can only say: great. Just keep in mind that this does not work on all systems - but maybe that does not apply to the