[Bug 543566] Review Request: tsocks - Library to allow transparent SOCKS proxying

[bugzilla at redhat.com]

Please do not reply directly to this email. All additional
comments should be made in the comments box of this bug.


https://bugzilla.redhat.com/show_bug.cgi?id=543566





--- Comment #4 from Michael Schwendt <mschwendt at gmail.com>  2009-12-11 16:50:34 EDT ---
> tsocks.i686: W: no-soname /usr/lib/libtsocks.so.1.8
> I checked to solve this one, but I'm not sure how to do it
> cleanly and what it affect. Can you enlight me on this?

libtsocks isn't an ordinary shared library, which would be linked with
applications at build-time and which would create a dependency on the internal
library DT_SONAME in the application (as well as corresponding library
dependencies in the RPM packages).  libtsocks is loaded manually via LD_PRELOAD
before the run-time linker loads other libraries needed for an application. 
Hence it doesn't need a SONAME (man ld). 

Which also explains why I questioned whether this is a library or an
application? ;)  To copy a more elaborate description from tsocks.8 and
extending it with a note on the preloading would be added value.

(One could argue whether the versioned files libtsocks.so.1.8 and
libtsocks.so.1 need to be present at all, since nothing uses them currently.
It's strange that the author installs the library this way instead of
installing the library only as a single libtsocks.so file. However, in parts of
the documentation he refers to the versioned library file names. Inconsistent.
Dropping them would need a change of the docs, too. Not worthwhile IMO.)


> - tsocks.i686: W: shared-lib-calls-exit /usr/lib/libtsocks.so.1.8
> exit at GLIBC_2.0
> I could patch the program, but I'm not sure it's worth it.

Assuming you have ran "rpmlint -i" on the binary rpm, yes, in this case it
would not be worth it. The preloaded library only exits in fatal OOM
conditions.


> - tsocks.i686: W: devel-file-in-non-devel-package /usr/lib/libtsocks.so
> tsocks need this file and it's not a devel-file.

True. It's a false positive for this particular package. It's important to
understand this rpmlint warning, though, to apply it correctly in the context
of ordinary shared library packages.


> * I submited scratch build for f11 and f12 on i386, x86_64, ppc and
> ppc64 and everything is good :
> http://koji.fedoraproject.org/koji/taskinfo?taskID=1851702
> http://koji.fedoraproject.org/koji/taskinfo?taskID=1851708

The 64-bit builds are bad, however. On x86_64 and ppc64, %{_libdir} expands to
/usr/lib64 (to allow for multi-lib installations), but /usr/bin/tsocks contains
a hardcoded /usr/lib where it won't find the library.

-- 
Configure bugmail: https://bugzilla.redhat.com/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.