I think the issue is that those values exceeded the maximum to manage as date/time values, thousandths of a second.
And to round 0,9995 implies a +1 second what doesn't seem properly to transform the text in a date/time value.
Preserving in the text only three decimals, conversion it's done properly.

I don't know if it can be considerer really a bug.

I agree about the likely cause but I do see it as a bug or needful improvement; an user could argue that if "12:23:34.9994" is a legitimate time, so is "12:23:34.9996".

I think I can fix it; I'd either:
 - parse the whole "SS.0000" bit as a float respecting the regional settings,
 - add fractional_seconds/length(fractional_seconds) to the ongoing value, or
 - if the precision must be limited to 1 ms, allow 1 s to be carried over to the seconds part.

Any hints on the relevant bit of code? I'll look into setting up a development environment. Also, is there a spec that says that decimal digits beyond the third must be truncated or rounded in conversion? I thought dates were floating-point numbers.

Oops, I forgot to mention that at work, this failed only when opening the text file directly or using 'Data/Text to Columns', but 'Edit/Paste Special' works OK. At home (also 4.3.0.4, but on 64-bit Ubuntu), only File/Open seems to fail. I'll double-check tomorrow.

(In reply to ariel cornejo from comment #0)
> Open the attached file with Calc; use Tab as a delimiter and choose "Date
> (YMD)" for either column. The values with at least 0.9995 fractional seconds
> will fail to be converted.

Unable to reproduce under GNU/Linux using v4.3.2.2. The initially imported values are displayed using a date-time format code of DD/MM/YY HH.MM however editing this to DD/MM/YYYY HH.MM.SS.0000000 reveals sub-second detail. Perhaps try highlighting the imported data and using the indicated format to see if it makes any difference?

Status set to NEEDINFO. Please set back to UNCONFIRMED once the requested information is provided.

(In reply to Owen Genat from comment #4)
> Unable to reproduce under GNU/Linux using v4.3.2.2. The initially imported
> values are displayed using a date-time format code of DD/MM/YY HH.MM however
> editing this to DD/MM/YYYY HH.MM.SS.0000000 reveals sub-second detail.
> Perhaps try highlighting the imported data and using the indicated format to
> see if it makes any difference?

I just tried with version 4.4.0.0.alpha0 (24fb87501ef9d5aa715d572de7eb5efe49a0d9c3). Values in rows 4,14,15,25,26,36-40 are imported as text:

* the cells show e.g. "2014-08-26 15:01:52.999609000" and are left aligned
* TYPE() evaluates to 2
* the format can't be changed
* interestingly, arithmetic expressions like =F2+1/24 work on these unconverted cells. Apparently they're being converted automatically.

These are the results of three conversion methods:

File > Open > Column type = Date (YMD):
Paste > Column type = Text > Text to columns > Column type = Date (YMD):
Cells with fractional part < 0.9995 are imported as dates and the rest remain as text. 

Paste > Column type = YMD:
OK. TYPE() evaluates to 1 and the format can be changed.

> Status set to NEEDINFO. Please set back to UNCONFIRMED once the requested
> information is provided.

I'm attaching a spreadsheet with the results of all three cases.

Any idea on the relevant bits of source for conversion? I built from source specifically to help tackle this.

Created attachment 107721 [details]
test results

(In reply to ariel cornejo from comment #5)
> I just tried with version 4.4.0.0.alpha0
> (24fb87501ef9d5aa715d572de7eb5efe49a0d9c3). Values in rows
> 4,14,15,25,26,36-40 are imported as text:

That is quite a strange result. I have no problem here using:

v4.1.6.2 Build ID: 40ff705089295be5be0aae9b15123f687c05b0a
v4.2.6.3 Build ID: 3fd416d4c6db7d3204c17ce57a1d70f6e531ee21
v4.3.2.2 Build ID: edfb5295ba211bd31ad47d0bad0118690f76407d
v4.4.0.0.alpha0+ Build ID: e21f6e3838a64f6c2517479d021e943e2ffcab94 TinderBox: Linux-rpm_deb-x86_64@46-TDF, Branch:master, Time: 2014-10-10_09:04:45

Even selecting "Date (YMD)" on the column is unnecessary. The "Detect special numbers" option seems to be sufficient to import all values as dates, which can then be formatted to display as required.

(In reply to Owen Genat from comment #7)
> (In reply to ariel cornejo from comment #5)
> > I just tried with version 4.4.0.0.alpha0
> > (24fb87501ef9d5aa715d572de7eb5efe49a0d9c3). Values in rows
> > 4,14,15,25,26,36-40 are imported as text:
> 
> That is quite a strange result. I have no problem here using:
> 
> v4.1.6.2 Build ID: 40ff705089295be5be0aae9b15123f687c05b0a
> v4.2.6.3 Build ID: 3fd416d4c6db7d3204c17ce57a1d70f6e531ee21
> v4.3.2.2 Build ID: edfb5295ba211bd31ad47d0bad0118690f76407d
> v4.4.0.0.alpha0+ Build ID: e21f6e3838a64f6c2517479d021e943e2ffcab94
> TinderBox: Linux-rpm_deb-x86_64@46-TDF, Branch:master, Time:
> 2014-10-10_09:04:45
>
> Even selecting "Date (YMD)" on the column is unnecessary. The "Detect
> special numbers" option seems to be sufficient to import all values as
> dates, which can then be formatted to display as required.

I just tried "Detect special numbers" and it works just like you say. Did you tick it for the tests you describe above?

Anyway, it's not obvious that it should be used and the conversion otherwise definitely shouldn't depend on the *value* of the decimal part. Something is failing silently. I'm intrigued by m.a.riosv's suggestion in #1 but am clueless as to why the conversion works so differently depending on where it's called from.

Import from file/open -> I can reproduce [column type "Date (YMD)"] with LibreOffice 3.5.0 
Build ID: d6cde02, linux
and
Version: 4.4.3.0.0+
Build ID: 8106e522c4ea2ae4441ec571579a38eeb6d9af04
TinderBox: Linux-rpm_deb-x86_64@46-TDF, Branch:libreoffice-4-4, Time: 2015-03-13_12:39:32

Values in rows 4,14,15,25,26,36-40 are imported as text.

When I try to paste in linux, I get only limited import dialog where I can not select column type. Detect special numbers doesn't work in this case and all dates are imported as text.

In windows 7, LO 4.4.1  import from file/open or paste is the same - dialog where I can select column type "Date (YMD)". Values in rows 4,14,15,25,26,36-40 are imported as text.

Setting to new.

Looks works fine now.
Version: 6.5.0.0.alpha0+ (x64)
Build ID: 60e8941fd581bb06cbf6be62edb8c387e7c07812
CPU threads: 4; OS: Windows 10.0 Build 19035; UI render: default; VCL: win; 
Locale: es-ES (es_ES); UI-Language: en-US
Calc: CL