Bruno Haible <br...@clisp.org> writes: > Po Lu wrote: >> >> Looking at https://apilevels.com/, especially the column "Cumulative >> >> usage", >> >> it seems there are no users of Android API level 13 or older any more. >> > >> > My experience differs > > My experience agrees with https://apilevels.com/ : Most smartphones > built around 2011-2012 had a life span of 3-8 years. That is, their > hardware broke before 2020 (either the battery, or the plastic cable, > or something else).
That's hardly relevant when these devices are either built new or sold refurbished/lightly used. >> > --many E-Book readers are shipped with API 8 or 9, > > Do you mean: were shipped, or are still shipped today? They are shipped today. Here is one such listing for refurbished merchandise on an e-commerce platform with aftermarket Android 2.3 firmware: https://item.taobao.com/item.htm?id=564895754182 This link may decline to load outside mainland China, but there is ample evidence that these devices continue to be relevant elsewhere, as in these Reddit threads, where multiple respondents are either considering or have acquired this device or its successor: https://old.reddit.com/r/ereader/comments/1esc5gx/tolino_shine_1_gen_in_2024/ (Android 2.3) https://old.reddit.com/r/ereader/comments/1h4amnq/bought_tolino_shine_2/ (Android 4.0) >> I should advise you now that the breadth of Android systems supported by >> Emacs is non-negotiable as a matter of principle for myself > > Whereas on my side, what is non-negotiable is a reasonable approach to > the set of supported platforms. [...] > 1. I thought you were building one Emacs binary per version, only, > namely with Android API level 8. If you eliminate the use of sysinfo() > at compile time like this, it means that load average detection will > not work (i.e. produce 0,0,0) even on the newest Android version, > in your Emacs binaries. Is that what you intend? Or will you produce, > say, two binaries, one built with Android API level 8, and one with a > newer one, say, Android API level 24? The latter would then be able > to have working load average detection, while the former wouldn't. No, nearly a dozen binaries are produced for different ranges of API levels and architectures: 8, 9, 17-21, 21-24, 21-29, 24-29, and 29+. > 2. The line after '# if' is not correctly indented. > 3. The preprocessor expression ought to be simplified. > 4. The comment a line later is not updated. > 5. The patch lacks a ChangeLog entry or commit message. > > Whether you fix problem #1, is your choice. But if you want a patch included > in Gnulib, it needs to fix problems 2, 3, 4, 5. I'm aware. I was not requesting that it be incorporated verbatim. > And, please, send patches at attachment, not inline in the mail. Some mail > user agents mangle whitespace in a way that inline patches are not applicable. OK.
