git/list[1] front-page[2] threads[3] people[4] search[5] about
 

[PATCH 0/5, resend] fork/exec removal series

From
Johannes Sixt <johannes.sixt@telecom.at>
Date
Oct 3, 2007, 20:09 UTC
Message-ID
<1191442180-15905-1-git-send-email-johannes.sixt@telecom.at>
In-Reply-To
<200709302340.17644.johannes.sixt@telecom.at>

Here is a series of patches that removes a number fork/exec pairs. They are replaced by delegating to start_command/finish_command/run_command.

I've addressed all issues that were raised. However, concerning whether git_connect() shoud return NULL or not, I decided to stay with my earlier approach to return NULL for non-forking connections. Consequently, I had to remove all error checks because there is now no unique value that can indicate failure. The motivation is that the alternate approach (to always return non-NULL for any successful connection and NULL for failure) would benefit *only* libification; but *if* that ever happens, a major audit (and possibly) rewrite in the callers has to be done anyway because currently we are juggling, leaking, and double-closing file descriptors like mad - in error paths and normal paths.

Patch 2 depends on Patch 1; otherwise, there are no dependencies. The series goes on top of 'master'. Note that 'next' has an additional call of git_connect() in the new transport.c.

 builtin-archive.c |    8 +--
 cache.h           |    4 +-
 connect.c         |  128 +++++++++++++++++++++++++---------------------------
 convert.c         |   30 +++----------
 diff.c            |   38 +--------------
 fetch-pack.c      |   43 +++++------------
 peek-remote.c     |    8 +--
 send-pack.c       |    8 +--
 8 files changed, 96 insertions(+), 171 deletions(-)
-- Hannes
Previous: Johannes SixtNext: Johannes Sixt
Message 11 of 30 in “fork/exec removal series”
  1. 0/5 fork/exec removal seriesJohannes Sixt, Sep 30, 2007
  2. 1/5 Change git_connect() to return a struct child_process instead of a pid_t.Johannes Sixt, Sep 30, 2007
  3. 2/5 Use start_command() in git_connect() instead of explicit fork/exec.Johannes Sixt, Sep 30, 2007
  4. 3/5 Use start_command() to run the filter instead of explicit fork/exec.Johannes Sixt, Sep 30, 2007
  5. 4/5 Use run_command() to spawn external diff programs instead of fork/exec.Johannes Sixt, Sep 30, 2007
  6. 5/5 Use start_comand() in builtin-fetch-pack.c instead of explicit fork/exec.Johannes Sixt, Sep 30, 2007
  7. Johannes SchindelinSep 30, 2007
  8. Johannes SchindelinSep 30, 2007
  9. Junio C HamanoSep 30, 2007
  10. Johannes SixtSep 30, 2007
  11. 0/5 fork/exec removal seriesJohannes Sixt, Oct 3, 2007
  12. 1/5 Change git_connect() to return a struct child_process instead of a pid_t.Johannes Sixt, Oct 3, 2007
  13. 2/5 Use start_command() in git_connect() instead of explicit fork/exec.Johannes Sixt, Oct 3, 2007
  14. 3/5 Use start_command() to run the filter instead of explicit fork/exec.Johannes Sixt, Oct 3, 2007
  15. 4/5 Use run_command() to spawn external diff programs instead of fork/exec.Johannes Sixt, Oct 3, 2007
  16. 5/5 Use start_comand() in builtin-fetch-pack.c instead of explicit fork/exec.Johannes Sixt, Oct 3, 2007
  17. Junio C HamanoOct 4, 2007
  18. Johannes SixtOct 4, 2007
  19. Johannes SixtOct 4, 2007
  20. Johannes SixtOct 1, 2007
  21. Junio C HamanoOct 1, 2007
  22. Johannes SixtOct 1, 2007
  23. Junio C HamanoOct 2, 2007
  24. Johannes SchindelinSep 30, 2007
  25. Junio C HamanoSep 30, 2007
  26. Johannes SchindelinSep 30, 2007
  27. Johannes SixtSep 30, 2007
  28. Johannes SchindelinSep 30, 2007
  29. Johannes SixtOct 1, 2007
  30. David KastrupOct 1, 2007

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.