=encoding UTF-8 =head1 NAME debputy - Manifest style Debian-based package builder =head1 SYNOPSIS B [S> >] B [S< I<< command options >> >] B B [B<--apply-changes>] [B<--acceptable-migration-issues=>I[,I,...]] B B [B<--debputy-manifest=>I] B B B B [--auto-fix] B B [--no-auto-fix] B B B B B B B [B<--tcp|--ws> [B<--host> I] [B<--port> I]] =head1 DESCRIPTION The B program is a manifest style Debian-based package builder. This man page documents some of the user facing commands. If you are using B with a screen reader, consider setting the environment variable B to 1. This will make many B commands remove a lot of irrelevant visual rendering. Especially ASCII art style rendering that will just be annoying to listen to. Additionally, some output will be described with text to replace the visual rendering. =head2 Commands =over 4 =item check-manifest The B command will cause B to parse the manifest and examine it for obvious mistakes. Note that the command will I catch all mistakes as some problems can only be detected during a build. As an example, B can detect mistyped manifest variables but not typos in path names for installation rules. =item migrate-from-dh The B command will attempt to migrate the current package to B. For this command to be successful, it must be run from the root of an unpacked Debian source package and the package should be using the B sequencer. If you are looking to migrate to B from B, you may want to have a look at the B file (in the source root or in F). That document is a how-to guide that as extended advice on migration from B. The migration can rerun with newer version of B that might provide more features since the previous one even inside the same level of adoption. As an example, B added support for automatic relationship substvars. Re-running the migrator for an already migrated package can be used to detect any unnecessary explicit relationship substvars as unnecessary. The default migration target is based on existing features in the packaging where present. As an example, a build-dependency on B will make the migration tool only run migrators relevant for B (assuming no other markers are present). If the migration cannot identify any target markers, it has a built-in default target. The default target will change over time as B and the migrator mature. The B<--migration-target> option can be used to overrule this automatic detection. This can be useful both to expand on the migration level without performing any changes or to choose a non default initial migration level. Generally, the migration tool can take you from "less adoption" towards "more adoption" of B but not the inverse. As an example, migrating from B towards B is supposed, but the reverse is not. Use of version control is recommended for undoing transition if necessary. If any migrations involve creating a new or changing an existing F, the migration tool will first write a draft to F. From there, it may be renamed to F automatically depending on B<--apply-changes> or B<--no-apply-changes>. It supports the following options: =over 4 =item B<--migration-target> I Explicitly define how far the migration should go. This will override the detected or built-in default as the desired migration target. See L for details about the concrete levels of integration that can be used. =item B<--acceptable-migration-issues> I, B<--acceptable-migration-issues=ALL> The migration may detect unsupported features in the package. Some, but not all, of these can be reduced to a warning by passing B<--acceptable-migration-issues> with the name of the issue as provided in the error message. The special value B in all upper case will cause all issues that can be reduced to a warning to be reduced to a warning. This is only useful to reduce issues to warnings if you are reasonable sure that you can remove the feature or convert it to something that B supports. In some cases, it might be helpful to comment out the offending feature and re-run the migration rather than using B<--acceptable.migration-issues>. As an example, if a single line of F file is problematic, commenting it out will have B migrate the rest of the file for you leaving you only to manually migrate a single line. =item B<--apply-changes>, B<--no-act>, B<--no-apply-changes> These options decide whether the migration tool should perform destructive actions such as overwriting the existing B and deleting packaging files that have successfully migrated. The default is currently to not perform destructive actions as the migration tool does not detect version control systems. If support for detecting version control systems is added, this default may change. Note that the migration may replace B regardless of this option as that is its output for the updated draft manifest. The B<--no-act> is an alias of B<--no-apply-changes> to match the name that B commands use. =back =item reformat I<< Note: This subcommand needs optional dependencies to work from B or B >> Apply the formatting style on the packaging files. This is the same style that B would have applied if requested to reformat the files. The B tool reacts to having a B field in F from where you can pick a named style. The recommended style is B, which is named such to match the B code formatter for B (which imposes a style that evolves over time). For packages that does not have the B field, B will result to looking up the maintainer (and possibly co-maintainers) for known style preferences in its built-in configuration. If B can derive a style that all parties would agree too (or the team style for packaging teams), then that style will be used. At the time of introduction, the B style is similar to that of B, since that was one of the most common style according to L, But the style is expected to evolve over time and the two styles may diverge over time. The command accepts the following options: =over 4 =item B<--style=black> Override the package style and use the B style. Any auto-detection or B setting will be ignored. =item B<--auto-fix>, B<--no-auto-fix> Decide whether any changes should be fixed automatically. Either way, a difference (B) is displayed to stdout if any changes were detected. =item B<--linter-exit-code>, B<--no-linter-exit-code> There is a convention among linter tools to return a non-zero exit code for "issues". The B<--linter-exit-code> will enforce this behaviour while the B<--no-linter-exit-code> will disable it. The B program will use exit code B<2> for "issues" as a "linter exit code" when linting based exit codes are active. Not having a linter based exit code can be useful if you want to run the tool programmatically to perform the action and you only want the exit code to tell whether there was a problem providing the results. If you rely on the exit code, you are recommended to explicitly pass the relevant variant of the flag even if the current default matches your wishes. =item B<--unknown-or-unsupported-style-is-ok>, B<--missing-style-is-ok> By default, B will exit with an error when it cannot determine which style to use. This is generally what you want for "per package" CI or other lint checks to inform you that the reformatting will not work. However, for "cross package" pipelines, like the default Debian Salsa CI pipeline, having B automatically work when a style is set and doing nothing otherwise is preferable, since the pipeline can then provide a B job for all consumers without worrying about breaking their pipelines. It can also be useful for scripts or automated mass-edits where you want B to fixup the changes you did if there is a known style without being hampered by the packages that have no known style. The B<--missing-style-is-ok> is a deprecated name since it does not correctly imply that unsupported styles are also considered ok. =item B<--supported-style-is-required> Exit with an error if no supported style can be found. This is the default behaviour but this option can be used to override settings to disable it. The error does not distinguish between no style found or an unsupported style found (both lead to an error). If you rely on the exit code, please set this option explicitly. =back =item lint I<< Note: This subcommand needs optional dependencies to work from B or B >> Run the linting tooling for Debian packaging files. This will run a linter to check the Debian packaging files. This command is useful for CI or for when you cannot use the language server feature. It provides the same diagnostics as B would but without requiring an LSP capable editor as intermediate. The output is only intended for human consumption. Machine readable is not a goal at this time. Note that at the time of writing, the B file is only B supported. If you have F, please also run B to get more thorough checks for that file for now. The B command will inform you about this issue in the output if a F file is detected. Some relevant options for this subcommand include: =over 4 =item B<--auto-fix> If B is aware of one "obvious" solution to the issue, just apply it. This will apply the changes directly to the file. Use of version control for the Debian packaging is recommended when using this option in case you want to undo the result. =item B<--spellcheck> Include spellchecking in the linting results. These are by default omitted, since they are slower and there are often false-positives. I: Currently, B<--auto-fix> with B<--spellcheck> will auto-correct all spelling mistakes with a single correction available. This can be suboptimal behaviour in some cases and therefore combing these options are not always recommended. =item B<--linter-exit-code>, B<--no-linter-exit-code> There is a convention among linter tools to return a non-zero exit code for "severe issues". The B<--linter-exit-code> will enforce this behaviour while the B<--no-linter-exit-code> will disable it. The B program will use exit code B<2> for "severe issue" as a "linter exit code" when linting based exit codes are active. Not having a linter based exit code can be useful if you want to run the tool programmatically to display the results and you only want the exit code to tell whether there was a problem providing the results. If you rely on the exit code, you are recommended to explicitly pass the relevant variant of the flag even if the current default matches your wishes. =item B<--lint-report-format> I Choose the output format of the resulting report. The B report is a terminal output report. The B writes the output to an XML in a JUnit 4 format (should be compatible with the xunit2 family of JUnit4 style input). This is useful for GitLab CI pipelines to get the results imported via GitLab's B CI pipeline feature. =item B<--report-output> I For reports that generate file system artifacts, choose where the output should be placed. Ignored with a warning for formats that do not generate file system outputs such as the B report format. =item B<--warn-about-check-manifest>, B<--no-warn-about-check-manifest> Whether B should remind you that it has short comings in regards to validating the B file. The warning will only appear if the manifest file exists. That is, the options have no effect either way when the file is not present. The B<--no-warn-about-check-manifest> is mostly used for generic pipelines that separately call B to ensure that the manifest is also validated or for cases where the limitation is known and accepted. =back A short comparison of B vs. other tools: =over 4 =item B The language server feature from B provides an interactive and online version of the linting from B directly in any LSP capable editor with the proper glue configuration. The LSP feature gives you instant gratification, some additional editor-only features and interactive choices of available quickfixes. The "downside" of the B feature is that it requires a LSP capable editor and each editor has their own glue configuration. Since the B language server is new, almost no editor has built-in glue configuration meaning it has a steeper learning curve to get started. Additionally, some times you want the checks for CI checks or the current state of the package without having to open each file in an editor. Here B covers the same issues without the need for anything else. =item B The primary difference between the B linter and B is that B works on "binary" artifacts. Even the source checks of B checks the packaged version of the source rather than the files you are directly working. This means that you have to do a package "build" for lintian to spot any changes, which causes slow feedback loops. Additionally, B can provide feedback regardless of whether your package can currently build. Accordingly, you can get help and hints even for problems that would prevent a package build. By nature of how B works, you can only get hints from lintian on matters that does not break the package build. On the flip side, because B is checking the assembled artifacts, it can check for issues that are only visible after a package build. Additionally, B also checks for issues in the upstream sources to some extent. Checking upstream artifacts and the resulting Debian packages are not in scope for B as the B is intended to be a mirror of the language server diagnostics. In summary: Use B (or B) for short feedback loops. Use B for slower but more thorough checks on resulting packages. =item B The B has a broader scope than B. If you are a happy B user, odds are that B will not do a lot for you. Though, B might still be relevant as the language server provides additional editor related features. =back =item lsp server I<< Note: This subcommand needs optional dependencies to work from B or B >> Start the B language server (per B specification). Many modern editors can delegate language support to a B or indirectly via other features like supporting B (which in turn can delegate to a language server). The B tool provides one for many common packaging formats via the B subcommand for file formats such as B, B and B (DEP-5). You will often need some editor specific glue configuration to link a given file format or name to the language server. The B might provide an example glue snippet for your editor. In that glue configuration, you will need to provide a command. Often, B will suffice (using the stdio transport). See B for other integration options such as TCP (B<--tcp>) or websocket (B<--ws>) plus related supporting options. If you need to debug an issue with the language server, the TCP integration (B<--tcp>) can be quite helpful. In this mode, you run B in a terminal before starting your editor. This means you have direct and unfiltered access to the B command and its output. Remember to update your editor to use TCP integration rather than stdio integration (and remember to swap back when you are done). Check the B if you need a different bind address for the language server. If you can choose the language ID for a given file, you are recommended to use the file name relative to the source root (such as B). The service does account for some known variations such as B (as used by B from B) and B (as used by B). See B for a list of known language IDs along with their aliases. When properly set up, the language server will offer a variety of features such as completion suggestions, hover documentation, "as you type" diagnostics, quickfixes, etc. Please see B for the full list of features per format. That command will also help identify mandatory and optional dependencies for the B command. Note many of the features are subject to the editor supporting them, correct language IDs being passed to B, etc. Options for this subcommand =over 4 =item B<--ignore-language-ids> When provided, B will ignore any language ID that the editor provides for any file. Instead, B will only rely on the file name for determining how to interpret the file content. Since B supports multiple file formats, it is needs to know what kind of file it is working with. The editor is supposed to provide this via a "Language ID" attribute. This enables you as a user in the editor to override the file format and have proper editor support no matter the filename. Unfortunately, most Debian packaging files do not have a language ID assigned in the LSP specification, so editors either provide a custom language ID or no custom language ID at all (that is, an empty string). When the editor does not provide a language ID for file, B will since 0.1.25 automatically attempt to derive the language from the filename. With this option (introduced in 0.1.29), B will always derive the language from the filename even if the editor provided a language ID. This can be helpful if your editor is providing language IDs that B does not recognize. As an example, in B with B the language ID is derived from the name of the buffer's major mode. If you tried to use B with a major mode that B does not recognize then without this option, B would "silently" do nothing. With this option, it would have worked provided the filename matched B's expectation no matter the major mode. On the downside, B will not provide correct advice unless the paths matches F<< .../debian/I >>. This can provide issues with some setups where the debian directory is implicit such as some "packaging-only" repos or some editor scratch pads. =item B<--tcp> or B<--ws> By default, the B language server will use B for communication with the editor. These options provide either the TCP integration mode (B<--tcp>) or the websocket integration mode (B<--ws>). In this mode, the B<--host> and B<--port> options can be used to choose the bind address. These options are mutually exclusive. The B<--ws> option requires B Debian package. =item B<--host> I, B<--port> I With B<--tcp> or B<--ws>, these option determines the bind address. The default is 127.0.0.1 for host and 2087 for the port. In integration modes that does not need a bind address (such as the B mode), this option is ignored. =back =item lsp editor-config B Provide an example configuration glue for using the B with the given editor if known. The snippets are maintained on a basis effort basis for editors without built-in config glue for the B. Please file an issue (or a merge request) at L if a snippet needs to be updated, added or removed. =item lsp features List in a human readable format details about what language IDs are handled by the B along with what features are provided for each file format/language ID. =item tool-support These commands are intended for other tools to consume the output. Output is generally JSON by default or supported via B<--output-format=json>. =over 4 =item export-reference-data [DATASET] The B command export reference data. If provided, only the named dataset will be exported. Otherwise, all datasets will be exported. The reference data includes descriptions of the keywords used in the data set, which is helpful to understand the data. =item supports-tool-command Tests whether B knows about the named command. Returns successfully if known and unsuccessfully if not known. =item annotate-debian-directory The B command will make B scan the F directory for known packaging files and annotate them with information. Identifying known packaging files is done on a best effort basis and B has the following sources of information: =over 4 =item Data from plugins Any installed B plugin can provide data about known packaging files. Most of B's "built-in" rules are stored in the B or the B plugin. These are installed in F by default. If any of the data in there is wrong, please file a bug or a bug against the package providing the data (often B or B). If the plugin provides the relevant data, B will expose B and B, which are best-effort guesses for the file is installed (or where files listed in it will be installed). Please check the B and B for the file to see when these field are applicable (and which case it is). Note that some files can be matched multiple times. As an example F and F will generally appear once per package, because they are installed in each package. =item Dynamic data from L (via L>) Additionally, B will ask B to resolve all relevant helper commands and their relevant config snippets. This data will be cross referenced with the plugin provided data where possible. This will detect files that B (and its plugins) does not know about, but it cannot provide any additional information. This part requires B<< debhelper (>= 13.12~) >> to work fully. With older versions, the output will include an B attribute denoting that B returned non-zero. Additionally, with B<< debhelper (>= 13.16~) >> the command will also provide data about files associated with some B-commands not active with the current set of B addons. When B is missing a file, it is typically because the command that uses that config file is not introspectable. Typically, that can be fixed by patching the command to include a command line a la: # INTROSPECTABLE: CONFIG-FILES pkgfile(foo) Assuming the command uses B from L to look up the config file. Notable case that will likely not work is F where there is no B package in F but F calls B. This holds equally for all debhelper config files and related commands. Here, the resulting file (if detected at all) might be associated with the wrong package. =back =back =item plugin list [I] =item plugin show B B These commands provides access to features that are provided by plugins (Note: many B features are plugin provided, so these commands also covers a lot of "built-in" features). These commands will access features of all plugins B even if the current package will not activate all of these plugins. Unless otherwise stated, all output is intended to be human readable rather than machine readable. Formatting may change between any version. Many of the B subcommands also provide a csv format. Note this output is B intended for scripting as the output is subject to change - both in form of format availability and content. The csv output is intended as an aid to users of screen readers for which csv files are easier to deal with than visually rendered tables. If you need a stable format of some particular output, please file a feature request at L or via B. You can use B and B to see which topics are applicable for each subcommand. Noteworthy topics include: =over 4 =item plugins This topic provides a listing of all plugins that B is aware of. This topic can only used with B and not with B. =item pluggable-manifest-rules (aliases: pmr, p-m-r) The B manifest provides a number of places where the packager can provide a number of different rules such as B vs. B vs. B under B. These are called pluggable manifest rules and this topic provides insights to which rules are available where. When used with B, B will list all pluggable manifest rules available. When used with B, a rule name must be provided and B will then provide details about the rule. These details include attributes available (where applicable) and any reference documentation provided by the plugin. As an example, here is how you get the details about the install rule: debputy plugin show pluggable-manifest-rules install When a rule name is ambiguous, B will ask that you use B instead of just B. As an example: debputy plugin show pluggable-manifest-rules TransformationRule::remove debputy plugin show pluggable-manifest-rules DpkgMaintscriptHelperCommand::remove Note the type names (such as B) are currently an implementation detail and may change in the future. =item packager-provided-files (aliases: ppf, p-p-f) This topic provides details about all "packager provided files". Packager provided files can be put into F from where B will pick them up and install them somewhere in the package. While this command shows all possible variants (by their stems), the B topic will B real files matched. When used with B, B will list all the packager provided files that B knows about. When used with B, some additional details will be given. In a few cases, the packager provided file will be processed first (as an example F will be passed to B and the processed version will then be installed instead). =item used-packager-provided-files (aliases: uppf, u-p-p-f) This topic provides a list of all packager provided files used in this source package. This topic differs from B in that it only shows files in actual use whereas the other topic lists all known stems. The listing will potentially include files that B could have picked up, but will not do so during a package build because the relevant plugin is not explicitly requested (typically via a Build-Depends). These cases are placed in a separate table and will be clearly marked. This topic can only used with B and not with B. This topic only works when the command is run from the root of an unpacked Debian source package (as B needs to read F and scan the F directory). =item metadata-detectors This topic provides a listing of all "metadata detectors". These are plugin provided code snippets that scan the final form of the package and add substvars (for B), generate maintscript snippets, or/and declare triggers. This topic can only used with B and not with B. =item manifest-variables This topic covers B manifest variables. The listing will list the common manifest variables by default along with their values in source context (if possible). Some of the special case manifest variables are hidden by default (use B to see the filter options). When used with B, B will list the reference documentation (if provided by the plugin) related to the value along with a few other details. As implied above, this will only show B variables. Any manifest variables provided directly in the manifest is B covered by these commands. =item automatic-discard-rules This topic covers automatic discard rules, which are rules that automatically filters out (discards) sources from installation rules by default. The listing will list all the available automatic discard rules. The B command will show reference documentation and an example of what the rule reacts to (if these have been provided by the plugin). As an example: debputy plugin show automatic-discard-rules la-files =item type-mappings This topic cover type mappings that explains how some non-trivial types are interpreted. These covers types like B and B, which are used by other features such as pluggable manifest rules. When used with B, any plugin provided documentation and example inputs will be displayed for that rule. =back =item autopkgtest-test-runner The B command is intended to be used by B or from autopkgtests to run the tests of plugins in installed mode. =item internal-command This is for internal-only usage only. Any subcommand under B may disappear or change options between any release without any warning. =back =head1 GENERAL OPTIONS The following options general options or root level options are available. =over 4 =item B<-h>, B<--help> Print usage information and exits. The information printed depends on which subcommands appear prior to this option. =item B<--version> Prints version information and exists. Cannot be used with subcommands. =item B<-d>, B<--debug> Enable debug logging and raw stack traces on errors. Some warnings become errors as a consequence. =item B<--no-pager> Some subcommands will in their default output format pipe it to a pager to give you a more pleasant experience if standard out is a terminal. Examples include many of the B commands. This option will disable the pager feature. Most option formats via B<--output-format> will imply B<--no-pager> as well for subcommands that support that option. Note: Assuming the environment has no pager configuration at all, B will use L with the B environment variable set to B<-FRMQSX>. Notable, the B<-F> option will cause B to immediately terminate if the output fits on the screen. =item B<--plugin> I This option causes I to be loaded as a part of the commands execution if the command needs to load plugin data. For commands that load all plugins by default, this option causes the command to fail if I cannot be loaded. For commands that are restrictive about which plugins are loaded, subcommand will load I in addition other plugins that would normally be loaded. The I can either be a plugin name or a filename. The B program considers parameter with a forward slash as a filename. Otherwise, the parameter is interpreted as a plugin name. When given a plugin name, B will search for the plugin in its plugin search path and load it from there. When given a file name, B will read that file as a plugin and use the basename minus any B<.json> or B<.json.in> extension as the plugin name. For packages that need a plugin that they provide themselves during their build process, this option can be useful to tell B about it. For the build itself, usually you want to use B. But this option can still be useful for B etc. The other use-case is to load a plugin not installed into the plugin search directories. Notably, you can use this to shadow an existing plugin, which can be useful for debugging and developing your plugin changes. This option cannot be used with bundled plugins. As an example, both B<--plugin debputy> and B<--plugin path/to/a/debputy.json> will result in an error. =item B<--debputy-manifest> F If the command needs to parse a manifest, have it read F instead of B. Note this is mostly for testing as other features might not work correctly if the manifest is not aligned with the current working directory. =back =head1 FILES =over 4 =item F Please see F for details on the format. If you are converting your first package, you may want to read F first. Unlike most debhelper like tools, this file is per source package rather than per binary package. Therefore, you I use F<< debian/I.debputy.manifest >> to provide a specialized manifest for I. Instead, all the needed parts should be written into the manifest itself. The B<--debputy-manifest> option can be used to have B process manifest other than F, which may be useful for testing or running B but not much else. =back =head1 INTEGRATION LEVELS The B has multiple levels of integrations, which defines how much of the packaging that B is handling relative to the default B sequence. The following integrations levels are available: =over 4 =item dh-sequence-zz-debputy-rrr This integration level replaces the minimal number of commands necessary to provide B support for B package (even those needing static ownership). The sequence is often compatible with other B sequences. To use this B integration level, any custom file ownership and mode I be migrated to the B. Custom binary package version (B<-v> to B) is supported via the manifest. This migration level corresponds to a B on B. The following debhelper commands are removed: =over 4 =item - dh_fixperms =item - dh_shlibdeps =item - dh_gencontrol =item - dh_md5sums =item - dh_builddeb =back Note the following B features are disabled in this integration mode: =over 4 =item - Installation rule (the B keyword in the manifest). Any installation of content that should go resulting B<.deb> or B<.udeb> should happen via B's mechanisms such as B. =item - Metadata detectors from plugins. Instead, substvars, maintscripts and triggers are handled and generated per B conventions. =back =item dh-sequence-zz-debputy With this integration level, B will take over all installation of files into the packages. This will replace basically all commands after the B command in the standard B sequence. This also makes the integration level incompatible with many debhelper add-ons, since they expect to run after B and assume contents will be materialized into F<< debian/I >>. This migration level corresponds to a B on B or B. =back =head1 SEE ALSO L =head1 AUTHOR Niels Thykier =cut