I decided the "algorithm" part was unmaintainable overkill, and I went back to the earlier version that was previously accepted by reviewers, that just reports the plain CLDR version, eg, "34.0" initially. Aside from a couple of cosmetic changes, the only change to the accepted version was that I now need a couple of lines in bootstrap_xlocale_private.h, since the commit "Allow bootstrapping localdef on non-FreeBSD systems". If there are no objections, I plan to commit this tomorrow.
- Queries
- All Stories
- Search
- Advanced Search
- Transactions
- Transaction Logs
Advanced Search
Nov 7 2020
Nov 1 2019
Rebased on top of the recent commit "tools/tools/locale: allow POSIX target to be built in parallel".
Oct 27 2019
Instead of "000", just use "0" as the algorithm version. So the value you get is "0:34.0" currently. No other changes. I'd like to commit this soon so that we can use it in PostgreSQL 13.
Oct 3 2019
Sorry for the long delay. Here is a new version that addresses a theoretical problem not handled before: what if our code changes for some reason, but the CLDR version doesn't? So I added an extra part to the version string. Changes:
Oct 8 2018
Fixes for style feedback from jhb. Thanks!
Oct 7 2018
Sep 15 2018
Instead of introducing a new function querylocaleversion(3), teach the existing querylocale(3) to return versions, as suggested by kib. I think I prefer it this way, but I'm not sure if it's more or less likely to survive contact with the standards process.
Sep 14 2018
In D17166#366077, @kib wrote:You could add a bit to the LC_XXX_MASK and make querylocale() returning either string or numeric representation of the version when the bit is set together with the real locale category bit.
Thank you!
strncpy->strlcpy, as suggested by yuripv.
Adjust man pages based on feedback from 0mp. While removing "All rights reserved", and after checking some nearby files and the documentation, it seemed that the thing to do with new files was to use my own name on the copyright message. I also added a new line to the querylocaleversion.3 page to mention the purpose of the version string.
New version because I missed one detail from yuripv's feedback (.Xr querylocaleversion 3).
Man page tidying + more context, based on feedback from yuripv.
Sep 13 2018
Aug 30 2018
Aug 15 2018
Thanks!
Jul 25 2018
In D16234#348439, @munro_ip9.org wrote:Would it make sense to check if "nm /path/to/postgres | grep setproctitle_fast"?
Jul 24 2018
In D16111#344337, @munro_ip9.org wrote:I'll commit the change to the PostgreSQL development branch soon, to affect PostgreSQL 12.
In D16234#344465, @girgen wrote:Looks good. We should probably also change the default configuration if setproctitle_fast exists. When was it introduced, in which OSVERSION?
files/patch-src_backend_utils_misc_postgresql.conf.sample turns setproctitle off.
Jul 12 2018
In D16234#344339, @mat wrote:Has this been submitted upstream?
In D16111#342204, @kib wrote:In D16111#342052, @munro_ip9.org wrote:Thanks kib! For the record, here's a patch for PostgreSQL that I plan to push when this makes it to a RELEASE:
Why waiting for a release ? There is a significant population of users utilizing stable, and even HEAD is sometimes used not only by developers. Early you push this into the wild world, more feedback you get.
Jul 9 2018
Added handling for EAGAIN on write() (not sure if that case is reachable).
Jul 7 2018
Jul 4 2018
Thanks kib! For the record, here's a patch for PostgreSQL that I plan to push when this makes it to a RELEASE:
Jul 3 2018
Updates from kib's code review.
May 12 2018
Apr 27 2018
Apr 19 2018
Thank you!
Apr 18 2018
Fixed assertion about struct layout on i386. Fixed typo in man page (s/pont/point/).
The "fork_no_inherit" and "exec_inherit" tests' child processes also need to call _exit(0), so here's a new version to do that. Thanks for the review jilles. I agree the tests look much better with those changes.
Used _exit(0) to exit the tests's subprocesses successfully. Use sigwait to wait for signals synchronously. Documentation tweaks.
Apr 17 2018
Use 'p' instead of td->td_proc in one more place that I'd missed. (Thanks to mjg for prod via IRC.) Sorry for the churn.
Thanks for the reviews and help!
Use 'p' variable instead of repeating td->td_proc, as requested.
Apr 16 2018
Updated documentation and moved PDEATHSIG handling into its own switch, as requested.