You are correct. I didn't look too well. They do buffered reads.
I did those as well some time ago (between 2.3.2 and 2.3.5) but undid
them because I suspected some unwanted side effects. Turned out the
side-effects weren't caused by the IO buffering, so maybe that change
should be reverted somehow....
Artem Bokhan wrote:
> That was strace -f -p 4676 2>&1 | grep -E "^read|^poll"
>
> Artem Bokhan пишет:
>> Paul J Stevens пишет:
>>> I looked at the code for smtp-sink.c.
>>>
>>> It does the same as dbmail-lmtp: read one char at a time. Only
>>> difference is that it uses getc(2) rather than read(2).
>>>
>> But it looks like it works with 4k blocks. It's 4096 times faster :)
>>
>> read(5, "AiAC0AIgA/AD8AIgBABC4AIgBfAC0AOw"..., 4096) = 4096
>> poll([{fd=5, events=POLLIN, revents=POLLIN}], 1, 0) = 1
>> poll([{fd=5, events=POLLIN, revents=POLLIN}], 1, 100000) = 1
>> read(5, "MDEIAAEx\r\nADEAMwA2ABAEMAAwADIAFA"..., 4096) = 4096
>> poll([{fd=5, events=POLLIN, revents=POLLIN}], 1, 0) = 1
>> poll([{fd=5, events=POLLIN, revents=POLLIN}], 1, 100000) = 1
>> read(5, "ICAgICANAAA4NjAwMjExMTAzMjM5CgAA"..., 4096) = 4096
>> poll([{fd=5, events=POLLIN, revents=POLLIN}], 1, 0) = 1
>> poll([{fd=5, events=POLLIN, revents=POLLIN}], 1, 100000) = 1
>
> _______________________________________________
> DBmail mailing list
> [email protected]
> https://mailman.fastxs.nl/mailman/listinfo/dbmail
>
--
________________________________________________________________
Paul Stevens paul at nfg.nl
NET FACILITIES GROUP GPG/PGP: 1024D/11F8CD31
The Netherlands________________________________http://www.nfg.nl
_______________________________________________
DBmail mailing list
[email protected]
https://mailman.fastxs.nl/mailman/listinfo/dbmail
