Thanks for the fixes; I installed the attached patch into the development version on GitHub. On 02/15/2018 02:05 PM, Michael H Deckers via tz wrote:
The following proposals contain the necessary corrections. I have also tried to indicate exactly which data are implied by the source. I see this as a small step to make the tzdb data traceable to its sources, and thus to clarify the uncertainty of tzdb data (when they are estimated, or derived from less reliable sources). The format of these indications can certainly be improved.
Thanks, though I'm afraid that the tzdata source doesn't lend itself well to this kind of detailed provenance annotation, partly because the data for a transition is partly at the end of one line and partly in the start of the next line, and there's little room for commentary that exactly matches the transition. I installed the attached patch, which uses a less-formal style to mention the 1911 Portuguese decree in every place that uses it. While we're on the topic of provenance, I use 'git blame' for tzdb provenance, and recommend it for doing meta-detective work about why data entries are the way they are. Although 'git blame' doesn't always give the info I want, it's convenient and doesn't get in the way of regular maintenance.
[pt 1911] specifies Cabo Verde to use UT - 02 h since 1912-01-01 02:00Z. Currently, tzdb has a switch to UT - 02 h in 1907 without source (Shanks by default). This switch would precede the one in [pt 1911] which appears to be unlikely (admittedly, we know from Guernsey on 1913-06-18 that such conclusions may be wrong).
For now let's go with your guess, as in the attached patch.
Note that Lisbon Mean Time as in [pt 1911] and in [Milne 1899] differs from the GMTOFF given here for it (from Shanks?).
The Shanks-derived offset was wrong, and was fixed in commit 2951fa3bbc8d68e894e96a2399322254e20714a2, first released in 2018a. I suppose you were looking at a pre-2018a copy of the relevant file.