<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd"> <html lang="en"> <head> <meta http-equiv="content-type" content="text/html; charset=utf-8"> <title>Development Notes</title> <link rel="stylesheet" type="text/css" href="mesa.css"> </head> <body> <div class="header"> <h1>The Mesa 3D Graphics Library</h1> </div> <iframe src="contents.html"></iframe> <div class="content"> <h1>Development Notes</h1> <ul> <li><a href="#style">Coding Style</a> <li><a href="#submitting">Submitting Patches</a> <li><a href="#release">Making a New Mesa Release</a> <li><a href="#extensions">Adding Extensions</a> </ul> <h2 id="style">Coding Style</h2> <p> Mesa is over 20 years old and the coding style has evolved over time. Some old parts use a style that's a bit out of date. If the guidelines below don't cover something, try following the format of existing, neighboring code. </p> <p> Basic formatting guidelines </p> <ul> <li>3-space indentation, no tabs. <li>Limit lines to 78 or fewer characters. The idea is to prevent line wrapping in 80-column editors and terminals. There are exceptions, such as if you're defining a large, static table of information. <li>Opening braces go on the same line as the if/for/while statement. For example: <pre> if (condition) { foo; } else { bar; } </pre> <li>Put a space before/after operators. For example, <tt>a = b + c;</tt> and not <tt>a=b+c;</tt> <li>This GNU indent command generally does the right thing for formatting: <pre> indent -br -i3 -npcs --no-tabs infile.c -o outfile.c </pre> <li>Use comments wherever you think it would be helpful for other developers. Several specific cases and style examples follow. Note that we roughly follow <a href="http://www.stack.nl/~dimitri/doxygen/">Doxygen</a> conventions. <br> <br> Single-line comments: <pre> /* null-out pointer to prevent dangling reference below */ bufferObj = NULL; </pre> Or, <pre> bufferObj = NULL; /* prevent dangling reference below */ </pre> Multi-line comment: <pre> /* If this is a new buffer object id, or one which was generated but * never used before, allocate a buffer object now. */ </pre> We try to quote the OpenGL specification where prudent: <pre> /* Page 38 of the PDF of the OpenGL ES 3.0 spec says: * * "An INVALID_OPERATION error is generated for any of the following * conditions: * * * <length> is zero." * * Additionally, page 94 of the PDF of the OpenGL 4.5 core spec * (30.10.2014) also says this, so it's no longer allowed for desktop GL, * either. */ </pre> Function comment example: <pre> /** * Create and initialize a new buffer object. Called via the * ctx->Driver.CreateObject() driver callback function. * \param name integer name of the object * \param type one of GL_FOO, GL_BAR, etc. * \return pointer to new object or NULL if error */ struct gl_object * _mesa_create_object(GLuint name, GLenum type) { /* function body */ } </pre> <li>Put the function return type and qualifiers on one line and the function name and parameters on the next, as seen above. This makes it easy to use <code>grep ^function_name dir/*</code> to find function definitions. Also, the opening brace goes on the next line by itself (see above.) <li>Function names follow various conventions depending on the type of function: <pre> glFooBar() - a public GL entry point (in glapi_dispatch.c) _mesa_FooBar() - the internal immediate mode function save_FooBar() - retained mode (display list) function in dlist.c foo_bar() - a static (private) function _mesa_foo_bar() - an internal non-static Mesa function </pre> <li>Constants, macros and enumerant names are ALL_UPPERCASE, with _ between words. <li>Mesa usually uses camel case for local variables (Ex: "localVarname") while gallium typically uses underscores (Ex: "local_var_name"). <li>Global variables are almost never used because Mesa should be thread-safe. <li>Booleans. Places that are not directly visible to the GL API should prefer the use of <tt>bool</tt>, <tt>true</tt>, and <tt>false</tt> over <tt>GLboolean</tt>, <tt>GL_TRUE</tt>, and <tt>GL_FALSE</tt>. In C code, this may mean that <tt>#include <stdbool.h></tt> needs to be added. The <tt>try_emit_</tt>* methods in src/mesa/program/ir_to_mesa.cpp and src/mesa/state_tracker/st_glsl_to_tgsi.cpp can serve as examples. </ul> <h2 id="submitting">Submitting patches</h2> <p> The basic guidelines for submitting patches are: </p> <ul> <li>Patches should be sufficiently tested before submitting. <li>Code patches should follow Mesa coding conventions. <li>Whenever possible, patches should only effect individual Mesa/Gallium components. <li>Patches should never introduce build breaks and should be bisectable (see <code>git bisect</code>.) <li>Patches should be properly formatted (see below). <li>Patches should be submitted to mesa-dev for review using <code>git send-email</code>. <li>Patches should not mix code changes with code formatting changes (except, perhaps, in very trivial cases.) </ul> <h3>Patch formatting</h3> <p> The basic rules for patch formatting are: </p> <ul> <li>Lines should be limited to 75 characters or less so that git logs displayed in 80-column terminals avoid line wrapping. Note that git log uses 4 spaces of indentation (4 + 75 < 80). <li>The first line should be a short, concise summary of the change prefixed with a module name. Examples: <pre> mesa: Add support for querying GL_VERTEX_ATTRIB_ARRAY_LONG gallium: add PIPE_CAP_DEVICE_RESET_STATUS_QUERY i965: Fix missing type in local variable declaration. </pre> <li>Subsequent patch comments should describe the change in more detail, if needed. For example: <pre> i965: Remove end-of-thread SEND alignment code. This was present in Eric's initial implementation of the compaction code for Sandybridge (commit 077d01b6). There is no documentation saying this is necessary, and removing it causes no regressions in piglit on any platform. </pre> <li>A "Signed-off-by:" line is not required, but not discouraged either. <li>If a patch address a bugzilla issue, that should be noted in the patch comment. For example: <pre> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=89689 </pre> <li>If there have been several revisions to a patch during the review process, they should be noted such as in this example: <pre> st/mesa: add ARB_texture_stencil8 support (v4) if we support stencil texturing, enable texture_stencil8 there is no requirement to support native S8 for this, the texture can be converted to x24s8 fine. v2: fold fixes from Marek in: a) put S8 last in the list b) fix renderable to always test for d/s renderable fixup the texture case to use a stencil only format for picking the format for the texture view. v3: hit fallback for getteximage v4: put s8 back in front, it shouldn't get picked now (Ilia) </pre> <li>If someone tested your patch, document it with a line like this: <pre> Tested-by: Joe Hacker <jhacker@foo.com> </pre> <li>If the patch was reviewed (usually the case) or acked by someone, that should be documented with: <pre> Reviewed-by: Joe Hacker <jhacker@foo.com> Acked-by: Joe Hacker <jhacker@foo.com> </pre> </ul> <h3>Testing Patches</h3> <p> It should go without saying that patches must be tested. In general, do whatever testing is prudent. </p> <p> You should always run the Mesa test suite before submitting patches. The test suite can be run using the 'make check' command. All tests must pass before patches will be accepted, this may mean you have to update the tests themselves. </p> <p> Whenever possible and applicable, test the patch with <a href="http://piglit.freedesktop.org">Piglit</a> to check for regressions. </p> <h3>Mailing Patches</h3> <p> Patches should be sent to the Mesa mailing list for review. When submitting a patch make sure to use git send-email rather than attaching patches to emails. Sending patches as attachments prevents people from being able to provide in-line review comments. </p> <p> When submitting follow-up patches you can use --in-reply-to to make v2, v3, etc patches show up as replies to the originals. This usually works well when you're sending out updates to individual patches (as opposed to re-sending the whole series). Using --in-reply-to makes it harder for reviewers to accidentally review old patches. </p> <p> When submitting follow-up patches you should also login to <a href="https://patchwork.freedesktop.org">patchwork</a> and change the state of your old patches to Superseded. </p> <h3>Reviewing Patches</h3> <p> When you've reviewed a patch on the mailing list, please be unambiguous about your review. That is, state either <pre> Reviewed-by: Joe Hacker <jhacker@foo.com> </pre> or <pre> Acked-by: Joe Hacker <jhacker@foo.com> </pre> Rather than saying just "LGTM" or "Seems OK". </p> <p> If small changes are suggested, it's OK to say something like: <pre> With the above fixes, Reviewed-by: Joe Hacker <jhacker@foo.com> </pre> which tells the patch author that the patch can be committed, as long as the issues are resolved first. </p> <h3>Marking a commit as a candidate for a stable branch</h3> <p> If you want a commit to be applied to a stable branch, you should add an appropriate note to the commit message. </p> <p> Here are some examples of such a note: </p> <ul> <li>CC: <mesa-stable@lists.freedesktop.org></li> <li>CC: "9.2 10.0" <mesa-stable@lists.freedesktop.org></li> <li>CC: "10.0" <mesa-stable@lists.freedesktop.org></li> </ul> Simply adding the CC to the mesa-stable list address is adequate to nominate the commit for the most-recently-created stable branch. It is only necessary to specify a specific branch name, (such as "9.2 10.0" or "10.0" in the examples above), if you want to nominate the commit for an older stable branch. And, as in these examples, you can nominate the commit for the older branch in addition to the more recent branch, or nominate the commit exclusively for the older branch. This "CC" syntax for patch nomination will cause patches to automatically be copied to the mesa-stable@ mailing list when you use "git send-email" to send patches to the mesa-dev@ mailing list. Also, if you realize that a commit should be nominated for the stable branch after it has already been committed, you can send a note directly to the mesa-stable@lists.freedesktop.org where the Mesa stable-branch maintainers will receive it. Be sure to mention the commit ID of the commit of interest (as it appears in the mesa master branch). The latest set of patches that have been nominated, accepted, or rejected for the upcoming stable release can always be seen on the <a href="http://cworth.org/~cworth/mesa-stable-queue/">Mesa Stable Queue</a> page. <h3>Criteria for accepting patches to the stable branch</h3> Mesa has a designated release manager for each stable branch, and the release manager is the only developer that should be pushing changes to these branches. Everyone else should simply nominate patches using the mechanism described above. The stable-release manager will work with the list of nominated patches, and for each patch that meets the crtieria below will cherry-pick the patch with: <code>git cherry-pick -x <commit></code>. The <code>-x</code> option is important so that the picked patch references the comit ID of the original patch. The stable-release manager may at times need to force-push changes to the stable branches, for example, to drop a previously-picked patch that was later identified as causing a regression). These force-pushes may cause changes to be lost from the stable branch if developers push things directly. Consider yourself warned. The stable-release manager is also given broad discretion in rejecting patches that have been nominated for the stable branch. The most basic rule is that the stable branch is for bug fixes only, (no new features, no regressions). Here is a non-exhaustive list of some reasons that a patch may be rejected: <ul> <li>Patch introduces a regression. Any reported build breakage or other regression caused by a particular patch, (game no longer work, piglit test changes from PASS to FAIL), is justification for rejecting a patch.</li> <li>Patch is too large, (say, larger than 100 lines)</li> <li>Patch is not a fix. For example, a commit that moves code around with no functional change should be rejected.</li> <li>Patch fix is not clearly described. For example, a commit message of only a single line, no description of the bug, no mention of bugzilla, etc.</li> <li>Patch has not obviously been reviewed, For example, the commit message has no Reviewed-by, Signed-off-by, nor Tested-by tags from anyone but the author.</li> <li>Patch has not already been merged to the master branch. As a rule, bug fixes should never be applied first to a stable branch. Patches should land first on the master branch and then be cherry-picked to a stable branch. (This is to avoid future releases causing regressions if the patch is not also applied to master.) The only things that might look like exceptions would be backports of patches from master that happen to look significantly different.</li> <li>Patch depends on too many other patches. Ideally, all stable-branch patches should be self-contained. It sometimes occurs that a single, logical bug-fix occurs as two separate patches on master, (such as an original patch, then a subsequent fix-up to that patch). In such a case, these two patches should be squashed into a single, self-contained patch for the stable branch. (Of course, if the squashing makes the patch too large, then that could be a reason to reject the patch.)</li> <li>Patch includes new feature development, not bug fixes. New OpenGL features, extensions, etc. should be applied to Mesa master and included in the next major release. Stable releases are intended only for bug fixes. Note: As an exception to this rule, the stable-release manager may accept hardware-enabling "features". For example, backports of new code to support a newly-developed hardware product can be accepted if they can be reasonably determined to not have effects on other hardware.</li> <li>Patch is a performance optimization. As a rule, performance patches are not candidates for the stable branch. The only exception might be a case where an application's performance was recently severely impacted so as to become unusable. The fix for this performance regression could then be considered for a stable branch. The optimization must also be non-controversial and the patches still need to meet the other criteria of being simple and self-contained</li> <li>Patch introduces a new failure mode (such as an assert). While the new assert might technically be correct, for example to make Mesa more conformant, this is not the kind of "bug fix" we want in a stable release. The potential problem here is that an OpenGL program that was previously working, (even if technically non-compliant with the specification), could stop working after this patch. So that would be a regression that is unaacceptable for the stable branch.</li> </ul> <h2 id="release">Making a New Mesa Release</h2> <p> These are the instructions for making a new Mesa release. </p> <h3>Get latest source files</h3> <p> Use git to get the latest Mesa files from the git repository, from whatever branch is relevant. This document uses the convention X.Y.Z for the release being created, which should be created from a branch named X.Y. </p> <h3>Perform basic testing</h3> <p> The release manager should, at the very least, test the code by compiling it, installing it, and running the latest piglit to ensure that no piglit tests have regressed since the previous release. </p> <p> The release manager should do this testing with at least one hardware driver, (say, whatever is contained in the local development machine), as well as on both Gallium and non-Gallium software drivers. The software testing can be performed by running piglit with the following environment-variable set: </p> <pre> LIBGL_ALWAYS_SOFTWARE=1 </pre> And Gallium vs. non-Gallium software drivers can be obtained by using the following configure flags on separate builds: <pre> --with-dri-drivers=swrast --with-gallium-drivers=swrast </pre> <p> Note: If both options are given in one build, both swrast_dri.so drivers will be compiled, but only one will be installed. The following command can be used to ensure the correct driver is being tested: </p> <pre> LIBGL_ALWAYS_SOFTWARE=1 glxinfo | grep "renderer string" </pre> If any regressions are found in this testing with piglit, stop here, and do not perform a release until regressions are fixed. <h3>Update version in file VERSION</h3> <p> Increment the version contained in the file VERSION at Mesa's top-level, then commit this change. </p> <h3>Create release notes for the new release</h3> <p> Create a new file docs/relnotes/X.Y.Z.html, (follow the style of the previous release notes). Note that the sha256sums section of the release notes should be empty at this point. </p> <p> Two scripts are available to help generate portions of the release notes: <pre> ./bin/bugzilla_mesa.sh ./bin/shortlog_mesa.sh </pre> <p> The first script identifies commits that reference bugzilla bugs and obtains the descriptions of those bugs from bugzilla. The second script generates a log of all commits. In both cases, HTML-formatted lists are printed to stdout to be included in the release notes. </p> <p> Commit these changes </p> <h3>Make the release archives, signatures, and the release tag</h3> <p> From inside the Mesa directory: <pre> ./autogen.sh make -j1 tarballs </pre> <p> After the tarballs are created, the sha256 checksums for the files will be computed and printed. These will be used in a step below. </p> <p> It's important at this point to also verify that the constructed tar file actually builds: </p> <pre> tar xjf MesaLib-X.Y.Z.tar.bz2 cd Mesa-X.Y.Z ./configure --enable-gallium-llvm make -j6 make install </pre> <p> Some touch testing should also be performed at this point, (run glxgears or more involved OpenGL programs against the installed Mesa). </p> <p> Create detached GPG signatures for each of the archive files created above: </p> <pre> gpg --sign --detach MesaLib-X.Y.Z.tar.gz gpg --sign --detach MesaLib-X.Y.Z.tar.bz2 gpg --sign --detach MesaLib-X.Y.Z.zip </pre> <p> Tag the commit used for the build: </p> <pre> git tag -s mesa-X.Y.X -m "Mesa X.Y.Z release" </pre> <p> Note: It would be nice to investigate and fix the issue that causes the tarballs target to fail with multiple build process, such as with "-j4". It would also be nice to incorporate all of the above commands into a single makefile target. And instead of a custom "tarballs" target, we should incorporate things into the standard "make dist" and "make distcheck" targets. </p> <h3>Add the sha256sums to the release notes</h3> <p> Edit docs/relnotes/X.Y.Z.html to add the sha256sums printed as part of "make tarballs" in the previous step. Commit this change. </p> <h3>Push all commits and the tag created above</h3> <p> This is the first step that cannot easily be undone. The release is going forward from this point: </p> <pre> git push origin X.Y --tags </pre> <h3>Install the release files and signatures on the distribution server</h3> <p> The following commands can be used to copy the release archive files and signatures to the freedesktop.org server: </p> <pre> scp MesaLib-X.Y.Z* people.freedesktop.org: ssh people.freedesktop.org cd /srv/ftp.freedesktop.org/pub/mesa mkdir X.Y.Z cd X.Y.Z mv ~/MesaLib-X.Y.Z* . </pre> <h3>Back on mesa master, add the new release notes into the tree</h3> <p> Something like the following steps will do the trick: </p> <pre> cp docs/relnotes/X.Y.Z.html /tmp git checkout master cp /tmp/X.Y.Z.html docs/relnotes git add docs/relnotes/X.Y.Z.html </pre> <p> Also, edit docs/relnotes.html to add a link to the new release notes, and edit docs/index.html to add a news entry. Then commit and push: </p> <pre> git commit -a -m "docs: Import X.Y.Z release notes, add news item." git push origin </pre> <h3>Update the mesa3d.org website</h3> <p> NOTE: The recent release managers have not been performing this step themselves, but leaving this to Brian Paul, (who has access to the sourceforge.net hosting for mesa3d.org). Brian is more than willing to grant the permission necessary to future release managers to do this step on their own. </p> <p> Update the web site by copying the docs/ directory's files to /home/users/b/br/brianp/mesa-www/htdocs/ with: <br> <code> sftp USERNAME,mesa3d@web.sourceforge.net </code> </p> <h3>Announce the release</h3> <p> Make an announcement on the mailing lists: <em>mesa-dev@lists.freedesktop.org</em>, and <em>mesa-announce@lists.freedesktop.org</em> Follow the template of previously-sent release announcements. The following command can be used to generate the log of changes to be included in the release announcement: <pre> git shortlog mesa-X.Y.Z-1..mesa-X.Y.Z </pre> </p> <h2 id="extensions">Adding Extensions</h2> <p> To add a new GL extension to Mesa you have to do at least the following. <ul> <li> If glext.h doesn't define the extension, edit include/GL/gl.h and add code like this: <pre> #ifndef GL_EXT_the_extension_name #define GL_EXT_the_extension_name 1 /* declare the new enum tokens */ /* prototype the new functions */ /* TYPEDEFS for the new functions */ #endif </pre> </li> <li> In the src/mapi/glapi/gen/ directory, add the new extension functions and enums to the gl_API.xml file. Then, a bunch of source files must be regenerated by executing the corresponding Python scripts. </li> <li> Add a new entry to the <code>gl_extensions</code> struct in mtypes.h </li> <li> Update the <code>extensions.c</code> file. </li> <li> From this point, the best way to proceed is to find another extension, similar to the new one, that's already implemented in Mesa and use it as an example. </li> <li> If the new extension adds new GL state, the functions in get.c, enable.c and attrib.c will most likely require new code. </li> <li> The dispatch tests check_table.cpp and dispatch_sanity.cpp should be updated with details about the new extensions functions. These tests are run using 'make check' </li> </ul> </div> </body> </html>