Is it OK to use DYLD_LIBRARY_PATH on Mac OS X? And, what’s the dynamic library search algorithm with it?

by

As you’ve noted, DYLD_LIBRARY_PATH behaves like LD_LIBRARY_PATH on other *nix. However, there is another environment variable you should look at called DYLD_FALLBACK_LIBRARY_PATH.

In general, these are (both on osx and linux) suggested only for development use as they can cause symbol lookup errors when you override with a library that does not have the same symbol table. A good example of this is when you attempt to override the default install of VecLib (e.g. blas lapack) with a custom install. This will cause symbol not found errors in applications linked to the system VecLib if DYLD_LIBRARY_PATH is set and the reverse (symbol lookup errors in custom applications) if it is not. This is due to the system blas/lapack not being a full implementation of the ATLAS libs.

DYLD_FALLBACK_LIBRARY_PATH will not produce these problems.

When installing libraries to a non-standard location, DYLD_FALLBACK_LIBRARY_PATH is much more sane. This will look for symbols in libraries provided in the default paths and if the symbol is not found there, fall back to the specified path.

The benefit is that this process will not cause symbol lookup errors in applications compiled against the default libraries.

In general, when libraries are installed to non-standard locations absolute paths should be specified which negates the ambiguity of the dynamic lookup.

More Related Contents:

  1. What are bitwise shift (bit-shift) operators and how do they work?
  2. GOTO still considered harmful? [closed]
  3. How should I do floating point comparison?
  4. What is an idempotent operation?
  5. How Does Modulus Divison Work
  6. What is the difference between concurrent programming and parallel programming?
  7. What does it mean to “program to an interface”?
  8. What is an existential type?
  9. What is name mangling, and how does it work?
  10. Given an integer, how do I find the next largest power of two using bit-twiddling?
  11. Why don’t languages raise errors on integer overflow by default?
  12. Under what circumstances are linked lists useful?
  13. Anemic Domain Model: Pros/Cons
  14. Is using a lot of static methods a bad thing?
  15. How to merge two files consistently line by line
  16. Why does division by zero in IEEE754 standard results in Infinite value?
  17. Fluent Interfaces – Method Chaining
  18. ICalendar and event updates not working in Outlook
  19. What is an ideal variable naming convention for loop variables? [closed]
  20. Should we compare floating point numbers for equality against a *relative* error?
  21. When is a do-while appropriate?
  22. How can I find the smallest difference between two angles around a point?
  23. Purity vs Referential transparency
  24. What does ‘foo’ really mean? [closed]
  25. Build an ASCII chart of the most commonly used words in a given text [closed]
  26. When a method has too many parameters? [closed]
  27. Converting EBNF to BNF
  28. Why are leading zeroes used to represent octal numbers?
  29. Print rpath of an executable on macOS
  30. What is differential execution?

Leave a Comment