This commit also removes the deprecated `Suffix` from MediaType. Now use `Suffixes` and put the MIME type suffix in the type, e.g. `application/svg+xml`.
9b494a58c Add comma in example config 1c24f672d Update multilingual.md 8bf85bfe4 Cosmetic change to test GitInfo 3b53b54f9 Update installing.md e3b8eb941 Get arolla-cocoon showcase ready for publishing 7b7931e59 Add "Cocoon Arolla showcase" c9cc7852c Add "scoop install hugo" for quick install on Windows 3037527fe Update Directory Structure following Hugo Pipes dbecb967d Add `assetDir` key to config list 488d1d034 Remove alias.txt
Before this commit, the live reload logic in `hugo server` got confused when you dropped a new bundle into the project while the server was running. The workaround was to restart the server.
This commit fixes the "live reload bundle detection" in server mode, and also makes sure that the bundle headers are always processed first.
hugolib: Use the interface value when doing Related search
Currently it makes no practical difference, but this is more a protection if we in the future creates index from the content related fields. That will not work from a shortcode.
tpl/tplimpl: Reimplement the ".Params tolower" template transformer
All `.Params` are stored lowercase, but it should work to access them `.Page.camelCase` etc. There was, however, some holes in the logic with the old transformer.
This commit fixes that by applying a blacklist instead of the old whitelist logic. `.Param` is a very distinct key. The original case will be kept in `.Data.Params.myParam`, but other than that it will be lowercased.
In Hugo 0.46 we made the output of what you get from resources.Get and similar static, i.e. language agnostic. This makes total sense, as it is wasteful and time-consuming to do SASS/SCSS/PostCSS processing for lots of languages when the output is lots of duplicates with different filenames.
But since we now output the result once only, this had a negative side effect for multihost setups: We publish the resource once only to the root folder (i.e. not to the language "domain folder").
This commit removes the language code from the processed image keys. This creates less duplication in the file cache, but it means that you should do a `hugo --gc` to clean up stale files.
d88477eb3 Fix some more redirect alias loops 209735670 Update features.md d4ec6ef87 Remove redirect loop from functions/default ce14da435 Fix the AlternativeOutputFormats/OutputFormats examples f64a22469 Update cond.md 357a70a32 Remove redirect loop from templates/404 4ad7b01ba Remove some aliases that were causing infinite redirect loops fd378e66a Fix alias redirect loop ec3629f6f Revert "Set code background to white" ffc56b880 Revert "Make inline code gray" 2a964e261 Make inline code gray 44069e09c Set code background to white d39c790d0 Linked my own blog2md migration script 3e55267be Release 0.46 7b6cfdd8b releaser: Prepare repository for 0.47-DEV d4cb54806 releaser: Add release notes to /docs for release of 0.46 6518ac7e7 releaser: Bump versions for release of 0.46 e54334d5c Merge commit '766085c2dc6fc95ac30fda2a9ebde2355fc12554'
81847981f Add notes for installing "extended" Sass/SCSS version 02da3bb58 Include usage of disqus internal template 9b53380c4 Update sitemap-template.md f3417dd52 Update sitemap-template.md aae33d9a8 Update sitemap-template.md 834edc205 Release 0.45.1 fb93ab905 Merge branch 'temp451' 79f350906 releaser: Prepare repository for 0.46-DEV 25cd2352c releaser: Add release notes to /docs for release of 0.45.1 e60377c71 releaser: Bump versions for release of 0.45.1 873f6dce2 Update features.md ecc116642 Update link for Azure cb88c3185 Added .Site.Home to Site Variiables (#541) 975a88791 Document includePaths 56c4e332f Release Hugo 0.45 c22b25210 Merge branch 'temp45' 0ffad3ada releaser: Prepare repository for 0.46-DEV 08d494cc3 releaser: Add release notes to /docs for release of 0.45 ad546ba45 releaser: Bump versions for release of 0.45 971c46bb2 Fixed "Sratch" to become "Scratch" on line 44 a28255bf3 Clarify the meaning of .Pages vs .Site.Pages 127aeee09 docs: Update ref, relref, GetPage docs 350d674ec resources.Concat doesn't have an alias 1fd1219b0 docs: Document refLinksErrorLevel and refLinksNotFoundURL af2252ff6 Merge commit 'a3535c8486b2ce762b1a8a9c30b03985c3e02cee' a11486805 Merge commit 'b6b37a1f00f808f3c0d2715f65ca2d3091f36495' 1c8896cb1 Fix addkit link to account for i18n 468aef3fc releaser: Prepare repository for 0.45-DEV c6f4b97a2 releaser: Add release notes to /docs for release of 0.44 d3985afb7 releaser: Bump versions for release of 0.44 2c59a330c Adjust release notes 83966769a releaser: Prepare repository for 0.44-DEV e539613f7 releaser: Add release notes to /docs for release of 0.43 2bf648944 releaser: Bump versions for release of 0.43 d9d6e4bf7 Fix typos d6798afda Merge commit '98293eaa1570b5aff4452021c8b6d6c8560b3f06' 37cc52261 Add a newScratch template func 1f7f09613 Merge branch 'release-0.42.2' 857b0b26a releaser: Prepare repository for 0.43-DEV
Make resources fetched via resources.Get and similar language agnostic
With the newly released Hugo Pipes, resources fetched and processed via `resources.Get` and similar was published to the relevant language sub folder when in multilingual mode.
The thought behind that was maximum flexibility with support for `assetDir` per language.
In practice this was a bad idea:
* You get duplication of identical content, with added processing time
* You end up with path issues that seem to be hard to find a way around (`@fa-font-path` is one example)
This commit changes that. Now there is only one `assetDir` and if you, as one example, need to generate a CSS per langugage, you need to set the paths yourself.
Rob Jackson [Tue, 31 Jul 2018 11:31:35 +0000 (12:31 +0100)]
Fix file paths for uncached transformed images
This commit also fixes an existing test to work according to the correct logic. The test was written based on erroneous behavior. We resize the image to 300x200px, and are now trying to fit it within a 50px square. The longest edge is 300 pixels, so we need to divide it by 6 (300 / 50 == 6). And then scale the shortest edge with the same proportion (200 / 6 == 33.33).
The original test was transforming the original source image, hence the previous values: