Re: [exim] 4.87 -> 4.88 readconf performance

2017-03-21 Thread Wayne
On Sat, Mar 18, 2017 at 05:49:42PM -0700, Rical Jasan wrote: > On 03/17/2017 10:46 AM, Wayne wrote: > >> Is there a reason to scan within comment blocks? > >> > > I assume based on the lack of responses that it's not expected that > > comment content be scanned at all? > > The developers would hav

Re: [exim] 4.87 -> 4.88 readconf performance

2017-03-18 Thread Rical Jasan
On 03/17/2017 10:46 AM, Wayne wrote: >> Is there a reason to scan within comment blocks? >> > I assume based on the lack of responses that it's not expected that > comment content be scanned at all? The developers would have to tell you if it is intentional, but as a user/administrator, I fully ex

Re: [exim] 4.87 -> 4.88 readconf performance

2017-03-17 Thread Wayne
On Tue, Mar 14, 2017 at 12:03:22PM -0400, Wayne wrote: > On Tue, Mar 14, 2017 at 02:15:56PM +0200, Lena--- via Exim-users wrote: > > > >> It triggers when a macro name contains the sequence /_[ODH]/. > > > > > > > > Does it include $header_Date: ? > > > > A workaround: $header_date: > > > > > > M

Re: [exim] 4.87 -> 4.88 readconf performance

2017-03-14 Thread Wayne
On Tue, Mar 14, 2017 at 02:15:56PM +0200, Lena--- via Exim-users wrote: > > >> It triggers when a macro name contains the sequence /_[ODH]/. > > > > > > Does it include $header_Date: ? > > > A workaround: $header_date: > > > > Macro, not header. > > Even comment lines are searched for /_[ODH]/ >

Re: [exim] 4.87 -> 4.88 readconf performance

2017-03-14 Thread Lena--- via Exim-users
> >> It triggers when a macro name contains the sequence /_[ODH]/. > > > > Does it include $header_Date: ? > > A workaround: $header_date: > > Macro, not header. Even comment lines are searched for /_[ODH]/ As far as I undrestand, every config line is searched before parsing. So, such line trigg

Re: [exim] 4.87 -> 4.88 readconf performance

2017-03-13 Thread Jeremy Harris
On 13/03/17 14:33, Lena--- via Exim-users wrote: >> It triggers when a macro name contains the sequence /_[ODH]/. > > Does it include $header_Date: ? > A workaround: $header_date: Macro, not header. -- J -- ## List details at https://lists.exim.org/mailman/listinfo/exim-users ## Exim details

Re: [exim] 4.87 -> 4.88 readconf performance

2017-03-13 Thread Lena--- via Exim-users
> This loop appears to be the problem. > It triggers when a macro name contains the sequence /_[ODH]/. Does it include $header_Date: ? A workaround: $header_date: > You are also correct in that when macros_create_builtin() is not > called the startup time is greatly reduced. -- ## List details

Re: [exim] 4.87 -> 4.88 readconf performance

2017-03-12 Thread Wayne
On Sun, Mar 12, 2017 at 11:24:45AM +, Jeremy Harris wrote: > With the 555 calls of macro_create() it does seem that the > builtin macros are being activated. The place that happens > is probably the call to macros_create_builtin() in > get_config_line() - about line 1029 of readconf.c in > mas

Re: [exim] 4.87 -> 4.88 readconf performance

2017-03-12 Thread Jeremy Harris
With the 555 calls of macro_create() it does seem that the builtin macros are being activated. The place that happens is probably the call to macros_create_builtin() in get_config_line() - about line 1029 of readconf.c in master HEAD. There's a commented-out printf on the line before; please unco

Re: [exim] 4.87 -> 4.88 readconf performance

2017-03-12 Thread lists-exim
On Thu, Mar 09, 2017 at 11:15:57AM -0600, René Berber via Exim-users wrote: > On 3/9/2017 10:46 AM, Jeremy Harris wrote: > > > On 09/03/17 16:20, lists-e...@useunix.net wrote: > >> Unfortunately this is stuck running this on HPUX 11.2 at the moment. > >> I don't have much installed aside from GCC/

Re: [exim] 4.87 -> 4.88 readconf performance

2017-03-10 Thread René Berber via Exim-users
On 3/9/2017 10:46 AM, Jeremy Harris wrote: > On 09/03/17 16:20, lists-e...@useunix.net wrote: >> Unfortunately this is stuck running this on HPUX 11.2 at the moment. >> I don't have much installed aside from GCC/GDB but if you can make >> some suggestions I'll be happy to do some research as time

Re: [exim] 4.87 -> 4.88 readconf performance

2017-03-09 Thread Jeremy Harris
On 09/03/17 16:20, lists-e...@useunix.net wrote: > Unfortunately this is stuck running this on HPUX 11.2 at the moment. > I don't have much installed aside from GCC/GDB but if you can make > some suggestions I'll be happy to do some research as time permits. > > I also plan to take this config and

Re: [exim] 4.87 -> 4.88 readconf performance

2017-03-09 Thread lists-exim
On Thu, Mar 09, 2017 at 09:10:13AM +, Jeremy Harris wrote: > On 09/03/17 02:16, lists-e...@useunix.net wrote: > > Most of the time is spent in the while loop in readconf_main around > > line 3424. All other portions of the function are negligible. > > That tells us it's reading the config... >

Re: [exim] 4.87 -> 4.88 readconf performance

2017-03-09 Thread Jeremy Harris
On 09/03/17 02:16, lists-e...@useunix.net wrote: > Most of the time is spent in the while loop in readconf_main around > line 3424. All other portions of the function are negligible. That tells us it's reading the config... Can you dig deeper? What tools do you have? -- Cheers, Jeremy -- ##

Re: [exim] 4.87 -> 4.88 readconf performance

2017-03-08 Thread lists-exim
On Mon, Mar 06, 2017 at 11:49:02AM -0500, lists-e...@useunix.net wrote: > On Sun, Mar 05, 2017 at 07:56:12PM +, Jeremy Harris wrote: > > On 04/03/17 22:27, Wayne Cuddy wrote: > > > Instrumenting the code with rudimentary timing metrics I found that on > > > average in versions 4.66-4.87 readcon

Re: [exim] 4.87 -> 4.88 readconf performance

2017-03-06 Thread lists-exim
On Sun, Mar 05, 2017 at 07:56:12PM +, Jeremy Harris wrote: > On 04/03/17 22:27, Wayne Cuddy wrote: > > Instrumenting the code with rudimentary timing metrics I found that on > > average in versions 4.66-4.87 readconf_main takes < 9000 > > microseconds. > > > > On 4.88 it takes on average 6300

Re: [exim] 4.87 -> 4.88 readconf performance

2017-03-05 Thread Jeremy Harris
On 04/03/17 22:27, Wayne Cuddy wrote: > Instrumenting the code with rudimentary timing metrics I found that on > average in versions 4.66-4.87 readconf_main takes < 9000 > microseconds. > > On 4.88 it takes on average 63000 microseconds. 4.88 added builtin macros (for assorted features) and a st

[exim] 4.87 -> 4.88 readconf performance

2017-03-05 Thread Wayne Cuddy
I've been running 4.66 on an internal HP-UX host for quite a long time. We've recently upgraded the hardware and the OS so I figure now is also a good time to upgrade EXIM to the latest. This is a relatively high volume host, 100+ msgs per day that relays mail to DMZ MX relays for internet des