# Tasks ## WONTFIX As it turns out, `/dev/urandom` is not POSIX either {#task-d1dc781f-8659-b02a-2b52-021528985f7c} - WONTFIX in 2021-07-15 Since POSIX doesn't have a [cryptographically secure pseudorandom number generator], I'll stick with the current solution, as there is no standard way to replace it. [cryptographically secure pseudorandom number generator]: https://en.wikipedia.org/wiki/Cryptographically-secure_pseudorandom_number_generator - INACTIVE in 2021-07-15 --- However I can't find a solution to this one. ## DONE As it turns out, `mktemp` is not POSIX {#task-f4807694-2187-5cb5-115e-c5970a052f1f} - DONE in 2021-07-15 Done in [`42354876de823f52a431eac900d360c074c6198d`](https://git.euandreh.xyz/git-permalink/commit/?id=42354876de823f52a431eac900d360c074c6198d). - TODO in 2021-07-15 ## DONE Fix empty HTML files for translated markdown in CI {#task-f957647c-df05-c0c1-a38b-2e6faea99b8f} - DONE in 2021-07-16 Done in [`870262a86860705d29ea9a2ce852c987b47add71`](https://git.euandreh.xyz/git-permalink/commit/?id=870262a86860705d29ea9a2ce852c987b47add71). - TODO in 2021-07-14 ## CANCELLED Also enforce translations in `assert-readme.sh` {#task-6946fc69-3f25-fc6b-5f4d-89cb940021f0} - CANCELLED in 2021-07-14 After an initial attempt, I realized that the generated markdown output depends on how mdpo emits the markdown, and that there are many ways to emit semantically equivalent markdown. One example is adding all the `[a link]: ...` to the bottom of the file instead or where it was placed initially. Since each README would have its own links, enforcing the translation becomes less feasible. - TODO in 2021-07-14 ## DONE Stop guessing languages in `assert-manpages.sh` {#task-be785104-8bf8-4107-349f-ae859c1db153} - DONE in 2021-07-14 Instead of guessing the language, `assert-manpages.sh` has a hardcoded list of supported languages, and now errors out when given a file with an unexpected naming or unsupported language. Done in [`92aa59b4481bfd30b9c29d3171c346e50fa93c9f`](https://git.euandreh.xyz/git-permalink/commit/?id=92aa59b4481bfd30b9c29d3171c346e50fa93c9f). Update: This was ultimately done and implemented in [`0f689438635d3a1c4cbf9cc1e62b69a72e642c68`](https://git.euandreh.xyz/git-permalink/commit/?id=0f689438635d3a1c4cbf9cc1e62b69a72e642c68). - TODO in 2021-07-14 ## DONE Fix link wrapping with md2po usage {#task-37fb45fb-d5d8-5511-2bd3-ebd2c3fee891} - DONE in 2021-07-16 Fixed upstream and in [`ae9b9b32aea1370478dc6d3daf7b769029d41bde`](https://git.euandreh.xyz/git-permalink/commit/?id=ae9b9b32aea1370478dc6d3daf7b769029d41bde). - WAITING in 2021-07-14 The issue filled upstream: . - TODO in 2021-07-14 There are still situations when the link gets broken with wrapping. This needs a little bit more time and investigation. - DONE in 2021-07-13 Fixed upstream in mdpo, done in [`a923f39376c511f128eb84a8b32b7464737a3972`](https://git.euandreh.xyz/package-repository/commit/?id=a923f39376c511f128eb84a8b32b7464737a3972). - TODO in 2021-07-13 ## DONE Add `installcheck` and `uninstallcheck` targets {#task-0b034315-cbd2-6fd6-fd32-9e00a12b7594} - DONE in 2021-07-07 Done in [`ded158891da92c8e402b408e934a49e5b00b5490`](https://git.euandreh.xyz/git-permalink/commit/?id=ded158891da92c8e402b408e934a49e5b00b5490). - TODO in 2021-07-07 ## CANCELLED Remove empty lines from roff source {#task-f65531fa-e44b-14b1-fa32-70a22346e060} - CANCELLED in 2021-07-14 Both pandoc and groff can understand empty lines in the roff source. Since there is no standard, it is reasonable to depend on implementation quirks. - TODO in 2021-07-07 --- From roff(7): > Never include empty or blank lines in a roff document. Instead, > use the empty request (a line consisting of a dot only) or a > line comment .\" if a structuring element is needed. ## DONE Support two categories of translations: mine and contributed {#task-b4d3eb41-64f4-9882-e03b-ab76af4dbd00} - DONE in 2021-07-15 Done in [`83e65d38d5269c7f505b5e65cd57ce640bba66bf`](https://git.euandreh.xyz/git-permalink/commit/?id=83e65d38d5269c7f505b5e65cd57ce640bba66bf). - TODO in 2021-07-06 --- For the translations that I write myself, I can do things such as give warnings and even enforce that they stay up-to-date. I can't do the same for contributed translation, both because I shouldn't enforce things, but also people may always leave. Having two tiers of translations would allow for contributions to fit in the workflow when available, be updated when possible, and included when requested, without creating trouble for the ones I maintain. Things to consider from the viewpoint of the contributor: - visibility when translations are missing; - how to contribute; - where to put translations; - where to get `.po` files from. ## DONE Investigate the accessibility and need for ARIA elements for manpages and markdown {#task-55fe6623-2bfc-49dd-061c-3899ae2b1c68} - DONE in 2021-07-13 As I found out, ARIA stands for "Accessible Rich Internet Applications", but since markdown and manpages include no JavaScript, the HTML itself being semantic is already accessible. This isn't even a "WONTFIX", but a "NONISSUE". - TODO in 2021-07-06 --- Eligible for investigations are: - manpages (I'd expect those to be standard, but who knows); - HTML from manpages as generated by pandoc; - HTML from markdown, also as generated by pandoc, both for `commonmark.sh` and `TODOs.sh`. Test them with a desktop screen reader, talkback and maybe a browser screen reader such as ChromeVox (or its free alternative). ## CANCELLED Add `EXIT_STATUS` section to manpage {#task-d6f1ddb3-2111-bcdb-1873-0a414a5f7157} - CANCELLED in 2021-06-27 This was already implemented :) In fact, this is a duplicate of [`#task-de34e119-049f-67a0-da8b-e0103ed24a2b`](#task-de34e119-049f-67a0-da8b-e0103ed24a2b). - TODO in 2021-06-27 ## DONE Translate README.md and CHANGELOG.md {#task-7c5cd2aa-6d92-0423-bfa7-81f2e8436586} - DONE in 2021-06-27 `README.md` translated in [`c7f999434f1299eaf6a4488380844f4205105b68`](https://git.euandreh.xyz/git-permalink/commit/?id=c7f999434f1299eaf6a4488380844f4205105b68). - DOING in 2021-06-27 `CHANGELOG.md` translated in [`cb6f339cf5bff6dec6b9c8b1bcf79496980ec36a`](https://git.euandreh.xyz/git-permalink/commit/?id=cb6f339cf5bff6dec6b9c8b1bcf79496980ec36a). - TODO in 2021-06-26 --- Unlike `doc/git-permalink.*.1.in` files, these two don't need to be commited into the repository. The reason why the manpages are is to lessen the requirement on the consumer when they're on the canonical `make && make install` path. They can even `make check` before installing (or uninstalling), and all of those targets are specifically kept with minimal requirements, so that the user can build, test and install the software without requiring extra dependencies. Due to that, the translated manpages are kept in the repository so that users don't need `po4a` or `gettext`. The same is not true for the `public` target. This is meant for building the HTML pages of the website. Even though all of the code that generates those pages live in the repository under `aux/`, `make public` isn't part of the canonical steps required for users to install software. Because of that, `{README,CHANGELOG}.{pt,fr,eo}.md` don't need to live in the repository, and can be generated on the fly under the `public` target. As with manpages, the actual file will be `README.en.md`, and `README.md` is a symlink to that. I should also add links to the translated versions on each translation. This should bring an accompanying change to `aux/workflow/assert-readme.sh` and `aux/workflow/assert-changelog.sh`. ## DONE Remove hard coded `public/` path in some files under `aux/` {#task-f09dedb7-1b25-0b5e-2520-910a9aa32562} - DONE in 2021-06-22 Done in [`57335b3e7c4c9e91c5747778180d611e5f15d144`](https://git.euandreh.xyz/git-permalink/commit/?id=57335b3e7c4c9e91c5747778180d611e5f15d144). - TODO in 2021-06-22 ## DONE Add test for install target {#task-755d8aa8-d39e-f8ce-d4c8-a8649338fffc} - DONE in 2021-06-22 Done in [`afecd9fc874922c01676cbb871a5159867a1ee67`](https://git.euandreh.xyz/git-permalink/commit/?id=afecd9fc874922c01676cbb871a5159867a1ee67). - TODO in 2021-06-22 ## DONE Use colours on `testing`/`test_ok` output {#task-cded56f2-c939-3c69-c818-0fb62d62cfd8} - DONE in 2021-06-22 Done in [`95a32d0d88783b880e67660865a0b3f69c705c25`](https://git.euandreh.xyz/git-permalink/commit/?id=95a32d0d88783b880e67660865a0b3f69c705c25). - TODO in 2021-06-22 ## DONE `--` doesn't work for `getopts` {#task-bebbe847-f552-be4b-b886-70a621162b69} - DONE in 2021-06-22 Done in [`298cfec38bd597b3c171f11d0d4451f6134243cf`](https://git.euandreh.xyz/git-permalink/commit/?id=298cfec38bd597b3c171f11d0d4451f6134243cf). - TODO in 2021-06-22 --- ```shell $ ./src/git-permalink.sh -p -- --version ./src/git-permalink.sh : option non permise -- - ./src/git-permalink.sh : option non permise -- v ./src/git-permalink.sh : option non permise -- e ./src/git-permalink.sh : option non permise -- r ./src/git-permalink.sh : option non permise -- s ./src/git-permalink.sh : option non permise -- i ./src/git-permalink.sh : option non permise -- o ./src/git-permalink.sh : option non permise -- n https://git.euandreh.xyz/git-permalink/tree/--version?id=5c2f15ac772615c390c0bc226a0f90ee37bad4e8 $ ./src/git-permalink.sh -p -- --help ./src/git-permalink.sh : option non permise -- - Usage: git permalink [-phV] FICHIER [LINENO] Options: -p seulement imprimez le lien, n'essayez pas de l'ouvrir -h, --help afficher ce message d'aide -V, --version imprime le numeró de version ``` ## DONE Add example to manpage on how to get the permalink of a file called `--help` {#task-ada9b39a-cc42-97b2-bd02-ad9c0adc475c} - DONE in 2021-06-22 Done in [`761a80b9e95c32b233d8f2a60208b7915795142b`](https://git.euandreh.xyz/git-permalink/commit/?id=761a80b9e95c32b233d8f2a60208b7915795142b). - TODO in 2021-06-22 ## DONE Write tests by creating and configuring repositories locally {#task-ab5a4835-3d79-03e4-8342-7f5add40b159} - DONE in 2021-06-22 Done in [`41683f99cc520c89e6bf5a15a2255e7928e461c6`](https://git.euandreh.xyz/git-permalink/commit/?id=41683f99cc520c89e6bf5a15a2255e7928e461c6). - TODO in 2021-06-22 ## CANCELLED Use `gettext` for translating `src/git-permalink.sh` texts {#task-717fa8b7-1c92-5766-8872-280ae061bf4b} - CANCELLED in 2021-06-21 I've considered this, and I even started working on it, but I've chosen not to pursue it further, as I now think this is a mistake. Regardless of gettext's power and usefulness, it is definitely an overkill for for this application. It would bring more things to the project, both in build time (requiring the generation of message files and friends) but also in run time (loading the message files themselves). This could certainly be done via `$PREFIX/libexec/$lang.sh` or something equivalent, but this would also be increasing the installation complexity, too. Expecting a sane installation environment is sane (but not always true), but the reasoning behind doing is just "because". The same functionality can be implemented with less, and it already is. It may be true that for bigger programs the same trade-off reasoning would tip the balance to the other side, and make choosing gettext a *better* decision, but this is not the case. For the modest purpose of this software, the current solution is good enough. Maybe it could be less restrictive of country or encoding, but the fundamental idea of embedded variables with an `eval` to define the relevant one stays the same. Changing from an `eval` to a `case` statement builds upon the same underlying principle: of having string variables on the same file, and keep them in sync in a completely manual process. Consider the cost and complexity of building tooling around the [quotation of the `don't` text][quotation], and one can envision one of the challenges that including in specialized i18n tooling might bring along. Those tools have their usages and places, they do exist for a good reason, and they should be used when applicable, but this is not it. This isn't really about POSIX sh, implementation language, or even the size of the program, but about the trade-off analysis of this use case, where I'm the one writing the code and the translations on a tiny piece of POSIX sh code. This analysis makes me choose **not** to use `gettext`. [quotation]: https://git.euandreh.xyz/git-permalink/tree/src/git-permalink.sh?id=caaac9ffb8c80432ef41ad9434762d3e3401f01a#n11 - TODO in 2021-06-21 ## DONE Make `OVERRIDES` and `SUPPORTED REMOTES` a subsection of `CONFIGURATION` {#task-9be7e4f5-df3c-ed5f-1ca1-2ce9fc61d944} - DONE in 2021-06-21 Done in [`c1f27ad0dadcab90939c157cd516a4204fd9919d`](https://git.euandreh.xyz/git-permalink/commit/?id=c1f27ad0dadcab90939c157cd516a4204fd9919d). - TODO in 2021-06-21 ## DONE Mark `[OPTIONS]` as optional in manpages {#task-1d80cea9-b246-0168-7844-34fe629b255f} - DONE in 2021-06-21 They're not really optional, but to be replaced by the option. Done in [`dd529f393357a1709fe2146f02720ff5fb215e6c`](https://git.euandreh.xyz/git-permalink/commit/?id=dd529f393357a1709fe2146f02720ff5fb215e6c). - TODO in 2021-06-21 --- Use `[\fIOPTIONS\fR]` in manpages. See `man man.1` for reference. ## DONE Add `EXIT STATUS` section in manpages {#task-de34e119-049f-67a0-da8b-e0103ed24a2b} - DONE in 2021-06-21 Done in [`1dff45eb075a8717a65c2a79e454b6e61c52f14d`](https://git.euandreh.xyz/git-permalink/commit/?id=1dff45eb075a8717a65c2a79e454b6e61c52f14d). - TODO in 2021-06-21 ## DONE Use `hunspell` for checking spelling mistakes in translations {#task-4ae91595-6e6d-be17-d882-754a478da878} - DONE in 2021-06-25 It is not included in the `dev-check` target because of errors of `iconv` and UTF-8: ```text $ ./aux/guix/with-container.sh 'make spellcheck' sh aux/workflow/assert-spelling.sh -l 'pt fr eo en' error - iconv: UTF-8 -> ANSI_X3.4-1968 error - iconv: UTF-8 -> ANSI_X3.4-1968 error - iconv: UTF-8 -> ANSI_X3.4-1968 error - iconv: UTF-8 -> ANSI_X3.4-1968 error - iconv: UTF-8 -> ANSI_X3.4-1968 error - iconv: UTF-8 -> ANSI_X3.4-1968 error - iconv: UTF-8 -> ANSI_X3.4-1968 error - iconv: ANSI_X3.4-1968 -> UTF-8 error - iconv: ANSI_X3.4-1968 -> UTF-8 error - iconv: ANSI_X3.4-1968 -> UTF-8 error - iconv: ANSI_X3.4-1968 -> UTF-8 error - iconv: ANSI_X3.4-1968 -> UTF-8 error - iconv: ANSI_X3.4-1968 -> UTF-8 error - iconv: ANSI_X3.4-1968 -> UTF-8 error - iconv: ANSI_X3.4-1968 -> UTF-8 error - iconv: ANSI_X3.4-1968 -> UTF-8 error - iconv: ANSI_X3.4-1968 -> UTF-8 error - iconv: ANSI_X3.4-1968 -> UTF-8 error - iconv: ANSI_X3.4-1968 -> UTF-8 public/fr/git-permalink.1.html:25: Locate: rer | Try: RER public/fr/git-permalink.1.html:29: Locate: remote | Try: remonte public/fr/git-permalink.1.html:29: Locate: origin | Try: origine public/fr/git-permalink.1.html:29: Locate: mod | Try: dom public/fr/git-permalink.1.html:30: Locate: apr | Try: par public/fr/git-permalink.1.html:34: Locate: sactiv | Try: activit public/fr/git-permalink.1.html:34: Locate: activit | Try: activait public/fr/git-permalink.1.html:48: Locate: RENCES | Try: RENDES public/fr/git-permalink.1.html:49: Locate: rences | Try: rendes public/fr/git-permalink.1.html:52: Locate: git | Try: g public/fr/git-permalink.1.html:52: Locate: berg | Try: ber public/fr/git-permalink.1.html:54: Locate: git | Try: g ``` This will be dealt with elsewhere (probably in my website's repository). The separate `spellcheck` target is enough. Done in [`f4deaf6bd66ed07cab897002c4cacd2e28027552`](https://git.euandreh.xyz/git-permalink/commit/?id=f4deaf6bd66ed07cab897002c4cacd2e28027552). - TODO in 2021-06-21 ## CANCELLED Consider `-?` flag for showing the help message {#task-fbc4e2cc-b9e0-53e2-6b0a-8ec118962046} - CANCELLED in 2021-06-21 It may have been true and conventional historically, but neither [POSIX], nor [TAOUP], nor [Conventions for Command Line Options] instruct that. [POSIX]: https://pubs.opengroup.org/onlinepubs/9699919799/basedefs/V1_chap12.html [TAOUP]: http://www.catb.org/~esr/writings/taoup/html/ch10s05.html [Conventions for Command Line Options]: https://nullprogram.com/blog/2020/08/01/ - TODO in 2021-06-21 ## DONE Use `getopts` in `aux/workflow/manpages.sh` {#task-9ee2bbc8-295f-52b7-4104-483869bad017} - DONE in 2021-06-23 Done in [`412a3f36cb40dba18f4f0f70a6804882ac6c68eb`](https://git.euandreh.xyz/git-permalink/commit/?id=412a3f36cb40dba18f4f0f70a6804882ac6c68eb). - TODO in 2021-06-22 Actually, I didn't implement this. What I did was add `getopts` to `src/git-permalink.sh.in`, not to `aux/workflow/manpages.sh`. - DONE in 2021-06-22 Done kind of in [`0ee2ce2c85d3376b35dc9338309e50e3c6f9eb80`](https://git.euandreh.xyz/git-permalink/commit/?id=0ee2ce2c85d3376b35dc9338309e50e3c6f9eb80). - TODO in 2021-06-21 --- Instead of the current confusing positional argument with some optional one. ## DONE Fix missing signature link {#task-eb65cbf1-bff9-be13-f185-3ea53333aa6f} - DONE in 2021-06-19 Done in [`3cd0c4f1f3798f5cf4b67ba1f6f36879bd6a033f`](https://git.euandreh.xyz/git-permalink/commit/?id=3cd0c4f1f3798f5cf4b67ba1f6f36879bd6a033f). - TODO in 2021-06-19 ## DONE Translate usage/help strings and manpages {#task-9e842529-b782-84a2-1d0a-801a8766cb7e} - DONE in 2021-06-19 Integration with Makefile done in [`b81c6437937fb594629d627334cc7e1fc2135507`](https://git.euandreh.xyz/git-permalink/commit/?id=b81c6437937fb594629d627334cc7e1fc2135507). - DOING in 2021-06-18 Translation of: - usage/help strings done in [`bf41672f72ee3ff7679c93e0d5a8f8646c19db3e`](https://git.euandreh.xyz/git-permalink/commit/?id=bf41672f72ee3ff7679c93e0d5a8f8646c19db3e); - manpages done in [`d0a5a24788e6c1720fce72629da4639fac372fad`](https://git.euandreh.xyz/git-permalink/commit/?id=d0a5a24788e6c1720fce72629da4639fac372fad). - TODO in 2021-06-16 ## DONE Fix tests in CI {#task-1e18a7cc-1055-bd09-3060-c03292ba583b} - DONE in 2021-06-12 There was a test that was too brittle, and it was removed. See [CI logs before] and [CI logs after]. Done in [`cd7bd39b1b6ce7d958f48d63b42ba3ab5fdf6f21`](https://git.euandreh.xyz/git-permalink/commit/?id=cd7bd39b1b6ce7d958f48d63b42ba3ab5fdf6f21). - TODO in 2021-06-12 --- For some reason, running them locally with `./aux/guix/with-container.sh 'make clean public dev-check'` passes, but the same command doesn't on CI. [CI logs before]: https://euandreh.xyz/git-permalink/ci-logs/2021-06-12T22:51:21+00:00-07a75e37a4564ed13bc2f50ab84009df137bc223.log [CI logs after]: https://euandreh.xyz/git-permalink/ci-logs/2021-06-12T23:15:22+00:00-cd7bd39b1b6ce7d958f48d63b42ba3ab5fdf6f21.log ## DONE Support override via git config {#task-7695d33d-e96e-d313-b74b-860addde01c3} - DONE in 2021-06-12 Done in [`dc51a628910eb872430e108084e81ee37f6e7dff`](https://git.euandreh.xyz/git-permalink/commit/?id=dc51a628910eb872430e108084e81ee37f6e7dff). - TODO in 2021-06-12 ## DONE Write `README.md`, `description` and `long-description` {#task-9bc160b8-5739-133d-1d46-61ba68ebd39e} - DONE in 2021-06-12 Finished `README.md` in [`d64e4491e417e31f35d3637ded05f5175c19b071`](https://git.euandreh.xyz/git-permalink/commit/?id=d64e4491e417e31f35d3637ded05f5175c19b071). - DOING in 2021-06-12 The `description` was already good enough. The `long-description` has better content over the previous duplication of whatever was in `description`. Done in [`9e089d29ec64936df39b9b1a0e385f988d843416`](https://git.euandreh.xyz/git-permalink/commit/?id=9e089d29ec64936df39b9b1a0e385f988d843416). The `README.md` is less empty, but still missing a "Usage" section. Done in [`a9022eb2f72efd9b99212a057df33cba762cd13b`](https://git.euandreh.xyz/git-permalink/commit/?id=a9022eb2f72efd9b99212a057df33cba762cd13b). - TODO in 2021-06-12 ## DONE Write tests {#task-fd654661-fa97-83db-1d49-83a66866ccfa} - DONE in 2021-06-12 I didn't find a simple way to test without getting too much into the weeds with mocking Git, and I don't think it is worth it. The tests I created are actually just exercising the help and versions flags, and asserting that other flags are ignored. For effectiveness this is pretty shallow and pretty useless, but I'm fine with the state of not testing the core functionality. To do that it would require me to split `src/git-permalink.sh` into two files, one with the function definitions and the other as a runner. But then I would have to glue them together on the `install` target, and more similar complications. Given all that, I chose not to write more tests, and call this done. Done in [`50a7c011274359ef058d30be87b9d29ca4fd06ed`](https://git.euandreh.xyz/git-permalink/commit/?id=50a7c011274359ef058d30be87b9d29ca4fd06ed). - TODO in 2021-06-12 ## DONE Write manpage {#task-6e6f917f-26ad-09fd-9a09-84b5df0f3ea4} - DONE in 2021-06-12 Done in [`0f02b1bf62005b451ff4b91a88716cb6e0365849`](https://git.euandreh.xyz/git-permalink/commit/?id=0f02b1bf62005b451ff4b91a88716cb6e0365849). - TODO in 2021-06-12 --- Show example of how to create an override URL on `git config` from a known URL. ## DONE Support default flags {#task-a66edd2e-cec4-ef7d-5a8a-3c9013381719} - DONE in 2021-06-12 Done in [`9aded035e3a6ab3fbac3129353db587b22132c79`](https://git.euandreh.xyz/git-permalink/commit/?id=9aded035e3a6ab3fbac3129353db587b22132c79). - TODO in 2021-06-12 --- Those are: - `-h` - `--help` - `-V` - `--version` ## DONE Support more forges {#task-cebc5298-17ad-5c60-dfa5-a25b66433a3a} - DONE in 2021-06-19 - TODO in 2021-06-12 --- The ones done so far as of this writing: - git.euandreh.xyz - sourcehut - savannah - gitlab - github The one I can think of right now but almost never use to merit being part of the first version: - Kernel CGit (where Git itself in publicly available): [`caaac9ffb8c80432ef41ad9434762d3e3401f01a`](https://git.euandreh.xyz/git-permalink/commit/?id=caaac9ffb8c80432ef41ad9434762d3e3401f01a) - codeberg: [`8be93bf6e7077d69eec43a80b773f9c2a6530524`](https://git.euandreh.xyz/git-permalink/commit/?id=8be93bf6e7077d69eec43a80b773f9c2a6530524) - notabug: [`0ee7b1fa813809e2e86e4e322999d3d56f1d180d`](https://git.euandreh.xyz/git-permalink/commit/?id=0ee7b1fa813809e2e86e4e322999d3d56f1d180d) - pagure: [`67573428bfcba5eb91cb286153ec593666c2a437`](https://git.euandreh.xyz/git-permalink/commit/?id=67573428bfcba5eb91cb286153ec593666c2a437) - bitbucket: [`2c5e8e11a26a9a09cbeb2ceda1ca7b4e4f3aa68a`](https://git.euandreh.xyz/git-permalink/commit/?id=2c5e8e11a26a9a09cbeb2ceda1ca7b4e4f3aa68a) - generic cgit? No, people should use `git-permalink.template-file-commit` and `git-permalink.template-commit-file` instead # Bugs # Improvements # Decisions ## DONE Add decision on using language identifier for translation, but not the country {#decision-ce5cd33e-b3a1-1606-0d78-0d9586b07529} - DONE in 2021-07-13 Decided! Using the language only is better, period. There are special cases where the country actually helps specify two different languages, such as zh_CN and zh_TW. But the default should be the language encompassing universally all countries that speak that language. Regionalisms are avoided to better inclusivity of less populated countries. BONUS: The golden rule is to make the translation unit to be the sentence rather than words. This is just a general comment on translation logistics, and has nothing to do with including the country. - TODO in 2021-07-06 --- Wikipedia has the best approach to translations, in my opinion, and I'm considering following it, or do something inspired by it. The main concept is do translate content based on the language, but not the country. So instead of having a `pt_PT` and a `pt_BR` version, there is only one `pt` version, and local differences are included in the translation itself. Other than the obvious benefit of requiring less translations to exist, I like the fact that they are capable of capturing the language, without excluding different countries or communities that speak them. Everywhere else I see the country as part of the language identifier or the translation, and I don't like it. Personally, I'd rather have a text that is a little bit less idiomatic to a country, but is universal to the language. Wikipedia has other decisions that I disagree with, such as serving different content for desktop and mobile, from different DNS addresses ([1], [2]). That is just to say that I want to take some of their ideas, but not all. [1]: https://diff.wikimedia.org/2009/06/30/wikimedia-mobile-launch/ [2]: https://www.mediawiki.org/wiki/Reading/Web/Mobile#About ## DONE Use `aux/workflow/manpages.sh` in the `make && make install` flow {#decision-6c2801da-bcb9-dd38-69bf-270a6f9b5acd} - DONE in 2021-06-23 - TODO in 2021-06-23 --- Excerpt from [`412a3f36cb40dba18f4f0f70a6804882ac6c68eb`](https://git.euandreh.xyz/git-permalink/commit/?id=412a3f36cb40dba18f4f0f70a6804882ac6c68eb) that implemented this change: ```text I disliked the fact that aux/workflow/manpages.sh still is being used for the "install" and "uninstall" targets. Before this file, the canonical workflow of "make && make install/uninstall" was 100% embedded within the Makefile itself. But now the Makefile calls to an external script for that. This isn't a real cost, other than how obvious the behaviour is for someone looking at the Makefile for the first time. I still chose to do it anyway, because there was already too many things in the Makefile itself, and it was getting worse with time. I made sure to never cross the line of relying on an external tool for the canonical "make && make install/uninstall", and even for "make check". Those all work without requiring any extra tool outside what POSIX defines, such as "sed", "awk", etc. Despite the cost of adding this detour from the liner Makefile flow, I found it to be worth it to call to the external script, as this script can now also be shared across projects, and the customized Makefile be made simpler. In other to remove the "-- $(do_subst)" horrendous hack, I chose to use an inference rule for ".in" files, and remove the "$(do_subst)" variable altogether. Now all the files that need to go through sed should end in ".in", and the Makefile will take care of producing it. The upside is that this model is much better integrater into make itself. ``` ## DONE Commit the generated manpage translations directly into the repository {#decision-1f7db3d5-002e-515e-f209-0d0917228359} - DONE in 2021-06-21 --- This is an anti-pattern in the viewpoint of the category of files that should exist in the repository, but this is made strictly for one rule: reducing the dependencies of the project for consumers (and not for developers), namely, gettext. Since gettext itself isn't standard (see [discussion] on this topic), adding this requirement to the consumer (user or packager) isn't desirable. So here I'm optimizing for more things being commited in the repository in the name of less requirements. [discussion]: https://www.austingroupbugs.net/view.php?id=1122 # Questions ## DONE Should I switch from pandoc to groff for generating HTML? {#question-539a1902-ed41-fc93-806a-088b6caf8134} - DONE in 2021-07-13 No. Pandoc is too lenient with man, and knows only about man and not about the rest of troffs toolchain (`tbl`, `eqn`, `pic`, etc.), but groff is worst on Unicode, so we stick to pandoc. - TODO in 2021-07-10 ## DONE Is an empty blank line in groff the same as a `.P` request? {#question-e3b7748d-2e67-b60b-e966-ad80c754ea58} - DONE in 2021-06-16 Done in [`c24b3a73b259943bb2440c587d8054ce90480552`](https://git.euandreh.xyz/git-permalink/commit/?id=c24b3a73b259943bb2440c587d8054ce90480552). - TODO in 2021-06-16 --- When reading the manpage with `man` or looking at the final output generated with `pandoc`, it seems to me that an empty line behaves and mean the same thing as a `.P` request. If this is true, these two snippets are equivalent: ```troff .\" First snippet, using .PP .PP The first paragraph. .PP The second paragrah. .PP The third paragraph. .\" Second snippet, using empty lines The first paragraph. The second paragrah. The third paragraph. ``` I find the second option to be much more readable, and if both are equivalent I prefer writing that. In fact, I don't see any reason at all to use `.PP`. # Resources # Scratch