/contrib/cvs/PROJECTS

https://bitbucket.org/freebsd/freebsd-head/ · #! · 53 lines · 41 code · 12 blank · 0 comment · 0 complexity · bab81b94c0dd8c63c938871804b3b49f MD5 · raw file

  1. This is a list of projects for CVS. In general, unlike the things in
  2. the TODO file, these need more analysis to determine if and how
  3. worthwhile each task is.
  4. I haven't gone through TODO, but it's likely that it has entries that
  5. are actually more appropriate for this list.
  6. 0. Improved Efficency
  7. * CVS uses a single doubly linked list/hash table data structure for
  8. all of its lists. Since the back links are only used for deleting
  9. list nodes it might be beneficial to use singly linked lists or a
  10. tree structure. Most likely, a single list implementation will not
  11. be appropriate for all uses.
  12. One easy change would be to remove the "type" field out of the list
  13. and node structures. I have found it to be of very little use when
  14. debugging, and each instance eats up a word of memory. This can add
  15. up and be a problem on memory-starved machines.
  16. Profiles have shown that on fast machines like the Alpha, fsortcmp()
  17. is one of the hot spots.
  18. * Dynamically allocated character strings are created, copied, and
  19. destroyed throughout CVS. The overhead of malloc()/strcpy()/free()
  20. needs to be measured. If significant, it could be minimized by using a
  21. reference counted string "class".
  22. * File modification time is stored as a character string. It might be
  23. worthwile to use a time_t internally if the time to convert a time_t
  24. (from struct stat) to a string is greater that the time to convert a
  25. ctime style string (from the entries file) to a time_t. time_t is
  26. an machine-dependant type (although it's pretty standard on UN*X
  27. systems), so we would have to have different conversion routines.
  28. Profiles show that both operations are called about the same number
  29. of times.
  30. * stat() is one of the largest performance bottlenecks on systems
  31. without the 4.4BSD filesystem. By spliting information out of
  32. the filesystem (perhaps the "rename database") we should be
  33. able to improve performance.
  34. * Parsing RCS files is very expensive. This might be unnecessary if
  35. RCS files are only used as containers for revisions, and tag,
  36. revision, and date information was available in easy to read
  37. (and modify) indexes. This becomes very apparent with files
  38. with several hundred revisions.
  39. 1. Improved testsuite/sanity check script
  40. * Need to use a code coverage tool to determine how much the sanity
  41. script tests, and fill in the holes.