[PATCH] emacs: derive correct timestamp in FCC unique name
On Wed, Jun 13 2012, Jesse Rosenthal wrote: > Previously, the timestamp at the beginning of the FCC maildir unique > maildir name was derived incorrectly, thanks to an integer > overflow. This changes the derivation of timestamp to float > arithmetic, and so gets the number correct. (It is still formatted > with "%d" so it will show up as an integer.) > > This change is mostly a question of consistency, since the unique name > is arbitrary anyway. But since most people use timestamps, and that was > the original intention here as well, we might as well. > > Signed-off-by: Jesse Rosenthal > --- Good point -- did some experiments (on 32bit machine, on 64 (* 65536 *65536) just works OK (4294967296). (insert (format " %d" (* 8191 65536))) 536805376 (insert (format " %d" (* 8192 65536))) -536870912 Wraps now (many times already) (insert (format " %d" (* 32767 65536.0))) 2147418112 (insert (format " %d" (* 32768 65536.0))) -2147483648 Wraps Tue Jan 19 03:14:08 2038 UTC (insert (format " %0.f" (* 32768 65536.0))) 2147483648 (insert (format " %0.f" (* 65536 65536.0))) 4294967296 Does not wrap. On emacs window you can re-experiment by typing C-x C-e after last closing ')' on line. (float-time '(32767 65536)) -> 2147418112.0 (minibuffer output) (float-time '(32768 65536)) -> error "Invalid time specification" On 64-bit machine this latest work ok; i.e 2147483648.0 is returned. Alternatives: 1) Use current patch, filenames will have extra '-' in 2038 on 32-bit systems. 2) Drop 'timeid' and replace it with (float-time) in `format` call a few lines in original source after the patch context below -- No idea how (float-time) works on 32-bit systems after 2038. 3) Use "%0.f" in format string instead of "%d" (there is no "%ld" or "%u" sequences for `format`. I wonder what (current-time-string) in current emacs return in 2038 (maybe I test (sometime(tm)) with (lib)faketime). I suggest option #2. Tomi > emacs/notmuch-maildir-fcc.el |2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/emacs/notmuch-maildir-fcc.el b/emacs/notmuch-maildir-fcc.el > index dcfbc4b..6fd8ff9 100644 > --- a/emacs/notmuch-maildir-fcc.el > +++ b/emacs/notmuch-maildir-fcc.el > @@ -141,7 +141,7 @@ will NOT be removed or replaced." > > (defun notmuch-maildir-fcc-make-uniq-maildir-id () > (let* ((ct (current-time)) > - (timeid (+ (* (car ct) 65536) (cadr ct))) > + (timeid (+ (* (car ct) 65536.0) (cadr ct))) > (microseconds (car (cdr (cdr ct > (hostname (notmuch-maildir-fcc-host-fixer system-name))) > (setq notmuch-maildir-fcc-count (+ notmuch-maildir-fcc-count 1)) > -- > 1.7.9.5
[PATCH] emacs: derive correct timestamp in FCC unique name
Hi, thanks for thinking this through. On Thu, 14 Jun 2012, Tomi Ollila wrote: > Alternatives: > > 1) Use current patch, filenames will have extra '-' in 2038 on 32-bit > systems. Well, that assumes there is still the same arithmetic operations -- the calendar issue will probably push them to either auto-convert to float or use bignum. But still, assuming that in 2038, people are still on 32bit machines, it seems we should minimize the amount of things that need to be fixed. So I agree that... > 2) Drop 'timeid' and replace it with (float-time) in `format` call a few > lines in original source after the patch context below -- No idea > how (float-time) works on 32-bit systems after 2038. is probably the best. They'll have to deal with their time_t (again, assuming the unlikely existence of 32bit machines then) but this doesn't depend on their arithemetic. I'll send a revision to this thread. > I suggest option #2. > > Tomi Thanks again, Jesse
Re: [PATCH] emacs: derive correct timestamp in FCC unique name
Hi, thanks for thinking this through. On Thu, 14 Jun 2012, Tomi Ollila wrote: > Alternatives: > > 1) Use current patch, filenames will have extra '-' in 2038 on 32-bit > systems. Well, that assumes there is still the same arithmetic operations -- the calendar issue will probably push them to either auto-convert to float or use bignum. But still, assuming that in 2038, people are still on 32bit machines, it seems we should minimize the amount of things that need to be fixed. So I agree that... > 2) Drop 'timeid' and replace it with (float-time) in `format` call a few > lines in original source after the patch context below -- No idea > how (float-time) works on 32-bit systems after 2038. is probably the best. They'll have to deal with their time_t (again, assuming the unlikely existence of 32bit machines then) but this doesn't depend on their arithemetic. I'll send a revision to this thread. > I suggest option #2. > > Tomi Thanks again, Jesse ___ notmuch mailing list notmuch@notmuchmail.org http://notmuchmail.org/mailman/listinfo/notmuch
Re: [PATCH] emacs: derive correct timestamp in FCC unique name
On Wed, Jun 13 2012, Jesse Rosenthal wrote: > Previously, the timestamp at the beginning of the FCC maildir unique > maildir name was derived incorrectly, thanks to an integer > overflow. This changes the derivation of timestamp to float > arithmetic, and so gets the number correct. (It is still formatted > with "%d" so it will show up as an integer.) > > This change is mostly a question of consistency, since the unique name > is arbitrary anyway. But since most people use timestamps, and that was > the original intention here as well, we might as well. > > Signed-off-by: Jesse Rosenthal > --- Good point -- did some experiments (on 32bit machine, on 64 (* 65536 *65536) just works OK (4294967296). (insert (format " %d" (* 8191 65536))) 536805376 (insert (format " %d" (* 8192 65536))) -536870912 Wraps now (many times already) (insert (format " %d" (* 32767 65536.0))) 2147418112 (insert (format " %d" (* 32768 65536.0))) -2147483648 Wraps Tue Jan 19 03:14:08 2038 UTC (insert (format " %0.f" (* 32768 65536.0))) 2147483648 (insert (format " %0.f" (* 65536 65536.0))) 4294967296 Does not wrap. On emacs window you can re-experiment by typing C-x C-e after last closing ')' on line. (float-time '(32767 65536)) -> 2147418112.0 (minibuffer output) (float-time '(32768 65536)) -> error "Invalid time specification" On 64-bit machine this latest work ok; i.e 2147483648.0 is returned. Alternatives: 1) Use current patch, filenames will have extra '-' in 2038 on 32-bit systems. 2) Drop 'timeid' and replace it with (float-time) in `format` call a few lines in original source after the patch context below -- No idea how (float-time) works on 32-bit systems after 2038. 3) Use "%0.f" in format string instead of "%d" (there is no "%ld" or "%u" sequences for `format`. I wonder what (current-time-string) in current emacs return in 2038 (maybe I test (sometime(tm)) with (lib)faketime). I suggest option #2. Tomi > emacs/notmuch-maildir-fcc.el |2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/emacs/notmuch-maildir-fcc.el b/emacs/notmuch-maildir-fcc.el > index dcfbc4b..6fd8ff9 100644 > --- a/emacs/notmuch-maildir-fcc.el > +++ b/emacs/notmuch-maildir-fcc.el > @@ -141,7 +141,7 @@ will NOT be removed or replaced." > > (defun notmuch-maildir-fcc-make-uniq-maildir-id () > (let* ((ct (current-time)) > - (timeid (+ (* (car ct) 65536) (cadr ct))) > + (timeid (+ (* (car ct) 65536.0) (cadr ct))) > (microseconds (car (cdr (cdr ct > (hostname (notmuch-maildir-fcc-host-fixer system-name))) > (setq notmuch-maildir-fcc-count (+ notmuch-maildir-fcc-count 1)) > -- > 1.7.9.5 ___ notmuch mailing list notmuch@notmuchmail.org http://notmuchmail.org/mailman/listinfo/notmuch
[PATCH] emacs: derive correct timestamp in FCC unique name
Previously, the timestamp at the beginning of the FCC maildir unique maildir name was derived incorrectly, thanks to an integer overflow. This changes the derivation of timestamp to float arithmetic, and so gets the number correct. (It is still formatted with "%d" so it will show up as an integer.) This change is mostly a question of consistency, since the unique name is arbitrary anyway. But since most people use timestamps, and that was the original intention here as well, we might as well. Signed-off-by: Jesse Rosenthal --- emacs/notmuch-maildir-fcc.el |2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/emacs/notmuch-maildir-fcc.el b/emacs/notmuch-maildir-fcc.el index dcfbc4b..6fd8ff9 100644 --- a/emacs/notmuch-maildir-fcc.el +++ b/emacs/notmuch-maildir-fcc.el @@ -141,7 +141,7 @@ will NOT be removed or replaced." (defun notmuch-maildir-fcc-make-uniq-maildir-id () (let* ((ct (current-time)) - (timeid (+ (* (car ct) 65536) (cadr ct))) + (timeid (+ (* (car ct) 65536.0) (cadr ct))) (microseconds (car (cdr (cdr ct (hostname (notmuch-maildir-fcc-host-fixer system-name))) (setq notmuch-maildir-fcc-count (+ notmuch-maildir-fcc-count 1)) -- 1.7.9.5
[PATCH] emacs: derive correct timestamp in FCC unique name
Previously, the timestamp at the beginning of the FCC maildir unique maildir name was derived incorrectly, thanks to an integer overflow. This changes the derivation of timestamp to float arithmetic, and so gets the number correct. (It is still formatted with "%d" so it will show up as an integer.) This change is mostly a question of consistency, since the unique name is arbitrary anyway. But since most people use timestamps, and that was the original intention here as well, we might as well. Signed-off-by: Jesse Rosenthal --- emacs/notmuch-maildir-fcc.el |2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/emacs/notmuch-maildir-fcc.el b/emacs/notmuch-maildir-fcc.el index dcfbc4b..6fd8ff9 100644 --- a/emacs/notmuch-maildir-fcc.el +++ b/emacs/notmuch-maildir-fcc.el @@ -141,7 +141,7 @@ will NOT be removed or replaced." (defun notmuch-maildir-fcc-make-uniq-maildir-id () (let* ((ct (current-time)) - (timeid (+ (* (car ct) 65536) (cadr ct))) + (timeid (+ (* (car ct) 65536.0) (cadr ct))) (microseconds (car (cdr (cdr ct (hostname (notmuch-maildir-fcc-host-fixer system-name))) (setq notmuch-maildir-fcc-count (+ notmuch-maildir-fcc-count 1)) -- 1.7.9.5 ___ notmuch mailing list notmuch@notmuchmail.org http://notmuchmail.org/mailman/listinfo/notmuch