Contents of /xml/htdocs/proj/en/glep/glep-0033.html

Parent Directory Parent Directory | Revision Log Revision Log

Revision 1.6 - (show annotations) (download) (as text)
Tue Oct 10 20:25:14 2006 UTC (12 years, 5 months ago) by g2boojum
Branch: MAIN
Changes since 1.5: +1 -1 lines
File MIME type: text/html
regenerate all .html files

1 <?xml version="1.0" encoding="utf-8" ?>
2 <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
3 <html xmlns="http://www.w3.org/1999/xhtml" xml:lang="en" lang="en">
4 <!--
5 This HTML is auto-generated. DO NOT EDIT THIS FILE! If you are writing a new
6 PEP, see http://www.python.org/peps/pep-0001.html for instructions and links
8 -->
9 <head>
10 <meta http-equiv="Content-Type" content="text/html; charset=utf-8" />
11 <meta name="generator" content="Docutils 0.4: http://docutils.sourceforge.net/" />
12 <title>GLEP 33 -- Eclass Restructure/Redesign</title>
13 <style type="text/css">
15 /*
16 :Author: David Goodger
17 :Contact: goodger@users.sourceforge.net
18 :date: $Date: 2003/06/02 17:03:08 $
19 :version: $Revision: 1.1 $
20 :copyright: This stylesheet has been placed in the public domain.
22 Default cascading style sheet for the PEP HTML output of Docutils.
23 */
25 .first {
26 margin-top: 0 }
28 .last {
29 margin-bottom: 0 }
31 .navigation {
32 width: 100% ;
33 background: #cc99ff ;
34 margin-top: 0px ;
35 margin-bottom: 0px }
37 .navigation .navicon {
38 width: 150px ;
39 height: 35px }
41 .navigation .textlinks {
42 padding-left: 1em ;
43 text-align: left }
45 .navigation td, .navigation th {
46 padding-left: 0em ;
47 padding-right: 0em ;
48 vertical-align: middle }
50 .rfc2822 {
51 margin-top: 0.5em ;
52 margin-left: 0.5em ;
53 margin-right: 0.5em ;
54 margin-bottom: 0em }
56 .rfc2822 td {
57 text-align: left }
59 .rfc2822 th.field-name {
60 text-align: right ;
61 font-family: sans-serif ;
62 padding-right: 0.5em ;
63 font-weight: bold ;
64 margin-bottom: 0em }
66 a.toc-backref {
67 text-decoration: none ;
68 color: black }
70 body {
71 margin: 0px ;
72 margin-bottom: 1em ;
73 padding: 0px }
75 dd {
76 margin-bottom: 0.5em }
78 div.section {
79 margin-left: 1em ;
80 margin-right: 1em ;
81 margin-bottom: 1.5em }
83 div.section div.section {
84 margin-left: 0em ;
85 margin-right: 0em ;
86 margin-top: 1.5em }
88 div.abstract {
89 margin: 2em 5em }
91 div.abstract p.topic-title {
92 font-weight: bold ;
93 text-align: center }
95 div.attention, div.caution, div.danger, div.error, div.hint,
96 div.important, div.note, div.tip, div.warning {
97 margin: 2em ;
98 border: medium outset ;
99 padding: 1em }
101 div.attention p.admonition-title, div.caution p.admonition-title,
102 div.danger p.admonition-title, div.error p.admonition-title,
103 div.warning p.admonition-title {
104 color: red ;
105 font-weight: bold ;
106 font-family: sans-serif }
108 div.hint p.admonition-title, div.important p.admonition-title,
109 div.note p.admonition-title, div.tip p.admonition-title {
110 font-weight: bold ;
111 font-family: sans-serif }
113 div.figure {
114 margin-left: 2em }
116 div.footer, div.header {
117 font-size: smaller }
119 div.footer {
120 margin-left: 1em ;
121 margin-right: 1em }
123 div.system-messages {
124 margin: 5em }
126 div.system-messages h1 {
127 color: red }
129 div.system-message {
130 border: medium outset ;
131 padding: 1em }
133 div.system-message p.system-message-title {
134 color: red ;
135 font-weight: bold }
137 div.topic {
138 margin: 2em }
140 h1 {
141 font-family: sans-serif ;
142 font-size: large }
144 h2 {
145 font-family: sans-serif ;
146 font-size: medium }
148 h3 {
149 font-family: sans-serif ;
150 font-size: small }
152 h4 {
153 font-family: sans-serif ;
154 font-style: italic ;
155 font-size: small }
157 h5 {
158 font-family: sans-serif;
159 font-size: x-small }
161 h6 {
162 font-family: sans-serif;
163 font-style: italic ;
164 font-size: x-small }
166 .section hr {
167 width: 75% }
169 ol.simple, ul.simple {
170 margin-bottom: 1em }
172 ol.arabic {
173 list-style: decimal }
175 ol.loweralpha {
176 list-style: lower-alpha }
178 ol.upperalpha {
179 list-style: upper-alpha }
181 ol.lowerroman {
182 list-style: lower-roman }
184 ol.upperroman {
185 list-style: upper-roman }
187 p.caption {
188 font-style: italic }
190 p.credits {
191 font-style: italic ;
192 font-size: smaller }
194 p.label {
195 white-space: nowrap }
197 p.topic-title {
198 font-family: sans-serif ;
199 font-weight: bold }
201 pre.line-block {
202 font-family: serif ;
203 font-size: 100% }
205 pre.literal-block, pre.doctest-block {
206 margin-left: 2em ;
207 margin-right: 2em ;
208 background-color: #eeeeee }
210 span.classifier {
211 font-family: sans-serif ;
212 font-style: oblique }
214 span.classifier-delimiter {
215 font-family: sans-serif ;
216 font-weight: bold }
218 span.interpreted {
219 font-family: sans-serif }
221 span.option-argument {
222 font-style: italic }
224 span.pre {
225 white-space: pre }
227 span.problematic {
228 color: red }
230 table {
231 margin-top: 0.5em ;
232 margin-bottom: 0.5em }
234 td, th {
235 padding-left: 0.5em ;
236 padding-right: 0.5em ;
237 vertical-align: top }
239 td.num {
240 text-align: right }
242 th.field-name {
243 font-weight: bold ;
244 text-align: left ;
245 white-space: nowrap }
247 h1 tt, h2 tt, h3 tt, h4 tt, h5 tt, h6 tt {
248 font-size: 100% }
250 tt {
251 background-color: #eeeeee }
253 ul.auto-toc {
254 list-style-type: none }
256 </style>
257 </head>
258 <body bgcolor="white">
259 <table class="navigation" cellpadding="0" cellspacing="0"
260 width="100%" border="0">
261 <tr><td class="navicon" width="150" height="35">
262 <a href="http://www.gentoo.org/" title="Gentoo Linux Home Page">
263 <img src="http://www.gentoo.org/images/gentoo-new.gif" alt="[Gentoo]"
264 border="0" width="150" height="35" /></a></td>
265 <td class="textlinks" align="left">
266 [<b><a href="http://www.gentoo.org/">Gentoo Linux Home</a></b>]
267 [<b><a href="http://www.gentoo.org/peps">GLEP Index</a></b>]
268 [<b><a href="http://www.gentoo.org/proj/en/glep/glep-0033.txt">GLEP Source</a></b>]
269 </td></tr></table>
270 <table class="rfc2822 docutils field-list" frame="void" rules="none">
271 <col class="field-name" />
272 <col class="field-body" />
273 <tbody valign="top">
274 <tr class="field"><th class="field-name">GLEP:</th><td class="field-body">33</td>
275 </tr>
276 <tr class="field"><th class="field-name">Title:</th><td class="field-body">Eclass Restructure/Redesign</td>
277 </tr>
278 <tr class="field"><th class="field-name">Version:</th><td class="field-body">1.6</td>
279 </tr>
280 <tr class="field"><th class="field-name">Last-Modified:</th><td class="field-body"><a class="reference" href="http://www.gentoo.org/cgi-bin/viewcvs.cgi/xml/htdocs/proj/en/glep/glep-0033.txt?cvsroot=gentoo">2006/09/05 20:54:30</a></td>
281 </tr>
282 <tr class="field"><th class="field-name">Author:</th><td class="field-body">Brian Harring &lt;ferringb&#32;&#97;t&#32;gentoo.org&gt;, John Mylchreest &lt;johnm&#32;&#97;t&#32;gentoo.org&gt;</td>
283 </tr>
284 <tr class="field"><th class="field-name">Status:</th><td class="field-body">Moribund</td>
285 </tr>
286 <tr class="field"><th class="field-name">Type:</th><td class="field-body">Standards Track</td>
287 </tr>
288 <tr class="field"><th class="field-name">Content-Type:</th><td class="field-body"><a class="reference" href="glep-0002.html">text/x-rst</a></td>
289 </tr>
290 <tr class="field"><th class="field-name">Created:</th><td class="field-body">29-Jan-2005</td>
291 </tr>
292 <tr class="field"><th class="field-name">Post-History:</th><td class="field-body">29-Jan-2005 6-Mar-2005 15-Sep-2005 5-Sep-2006</td>
293 </tr>
294 </tbody>
295 </table>
296 <hr />
297 <div class="contents topic">
298 <p class="topic-title first"><a id="contents" name="contents">Contents</a></p>
299 <ul class="simple">
300 <li><a class="reference" href="#status" id="id2" name="id2">Status</a></li>
301 <li><a class="reference" href="#abstract" id="id3" name="id3">Abstract</a></li>
302 <li><a class="reference" href="#terminology" id="id4" name="id4">Terminology</a></li>
303 <li><a class="reference" href="#motivation-and-rationale" id="id5" name="id5">Motivation and Rationale</a></li>
304 <li><a class="reference" href="#specification" id="id6" name="id6">Specification</a><ul>
305 <li><a class="reference" href="#ebuild-libraries-elibs-for-short" id="id7" name="id7">Ebuild Libraries (elibs for short)</a></li>
306 <li><a class="reference" href="#the-reduced-role-of-eclasses-and-a-clarification-of-existing-eclass-requirements" id="id8" name="id8">The reduced role of Eclasses, and a clarification of existing Eclass requirements</a></li>
307 <li><a class="reference" href="#the-end-of-backwards-compatibility" id="id9" name="id9">The end of backwards compatibility...</a></li>
308 <li><a class="reference" href="#tree-restructuring" id="id10" name="id10">Tree restructuring</a></li>
309 <li><a class="reference" href="#the-start-of-a-different-phase-of-backwards-compatibility" id="id11" name="id11">The start of a different phase of backwards compatibility</a></li>
310 <li><a class="reference" href="#migrating-to-the-new-setup" id="id12" name="id12">Migrating to the new setup</a></li>
311 </ul>
312 </li>
313 <li><a class="reference" href="#backwards-compatibility" id="id13" name="id13">Backwards Compatibility</a></li>
314 <li><a class="reference" href="#copyright" id="id14" name="id14">Copyright</a></li>
315 </ul>
316 </div>
317 <div class="section">
318 <h1><a class="toc-backref" href="#id2" id="status" name="status">Status</a></h1>
319 <p>Approved by the Gentoo Council on 15 September 2005. As of Sept. 2006
320 this GLEP is on hold, pending future revisions.</p>
321 </div>
322 <div class="section">
323 <h1><a class="toc-backref" href="#id3" id="abstract" name="abstract">Abstract</a></h1>
324 <p>For any design, the transition from theoretical to applied exposes inadequacies
325 in the original design. This document is intended to document, and propose a
326 revision of the current eclass setup to address current eclass inadequacies.</p>
327 <p>This document proposes several things- the creation of ebuild libraries, 'elibs',
328 a narrowing of the focus of eclasses, a move of eclasses w/in the tree, the
329 addition of changelogs, and a way to allow for simple eclass gpg signing.
330 In general, a large scale restructuring of what eclasses are and how they're
331 implemented. Essentially version two of the eclass setup.</p>
332 </div>
333 <div class="section">
334 <h1><a class="toc-backref" href="#id4" id="terminology" name="terminology">Terminology</a></h1>
335 <p>From this point on, the proposed eclass setup will be called 'new eclasses', the
336 existing crop (as of this writing) will be referenced as 'old eclasses'. The
337 distinction is elaborated on within this document.</p>
338 </div>
339 <div class="section">
340 <h1><a class="toc-backref" href="#id5" id="motivation-and-rationale" name="motivation-and-rationale">Motivation and Rationale</a></h1>
341 <p>Eclasses within the tree currently are a bit of a mess- they're forced to
342 maintain backwards compatibility w/ all previous functionality. In effect,
343 their api is constant, and can only be added to- never changing the existing
344 functionality. This obviously is quite limiting, and leads to cruft accruing in
345 eclasses as a eclasses design is refined. This needs to be dealt with prior to
346 eclass code reaching a critical mass where they become unmanageable/fragile
347 (recent pushes for eclass versioning could be interpreted as proof of this).</p>
348 <p>Beyond that, eclasses were originally intended as a method to allow for ebuilds
349 to use a pre-existing block of code, rather then having to duplicate the code in
350 each ebuild. This is a good thing, but there are ill effects that result from
351 the current design. Eclasses inherit other eclasses to get a single function- in
352 doing so, modifying the the exported 'template' (default src_compile, default
353 src_unpack, various vars, etc). All the eclass designer was after was reusing a
354 function, not making their eclass sensitive to changes in the template of the
355 eclass it's inheriting. The eclass designer -should- be aware of changes in the
356 function they're using, but shouldn't have to worry about their default src_*
357 and pkg_* functions being overwritten, let alone the env changes.</p>
358 <p>Addressing up front why a collection of eclass refinements are being rolled into
359 a single set of changes, parts of this proposal -could- be split into multiple
360 phases. Why do it though? It's simpler for developers to know that the first
361 eclass specification was this, and that the second specification is that,
362 rather then requiring them to be aware of what phase of eclass changes is in
363 progress.</p>
364 <p>By rolling all changes into one large change, a line is intentionally drawn in
365 the sand. Old eclasses allowed for this, behaved this way. New eclasses allow
366 for that, and behave this way. This should reduce misconceptions about what is
367 allowed/possible with eclasses, thus reducing bugs that result from said
368 misconceptions.</p>
369 <p>A few words on elibs- think of them as a clear definition between behavioral
370 functionality of an eclass, and the library functionality. Eclass's modify
371 template data, and are the basis for other ebuilds- elibs, however are <em>just</em>
372 common bash functionality.</p>
373 <p>Consider the majority of the portage bin/* scripts- these all are candidates for
374 being added to the tree as elibs, as is the bulk of eutils.</p>
375 </div>
376 <div class="section">
377 <h1><a class="toc-backref" href="#id6" id="specification" name="specification">Specification</a></h1>
378 <p>The various parts of this proposal are broken down into a set of changes and
379 elaborations on why a proposed change is preferable. It's advisable to the
380 reader that this be read serially, rather then jumping around.</p>
381 <div class="section">
382 <h2><a class="toc-backref" href="#id7" id="ebuild-libraries-elibs-for-short" name="ebuild-libraries-elibs-for-short">Ebuild Libraries (elibs for short)</a></h2>
383 <p>As briefly touched upon in Motivation and Rationale, the original eclass design
384 allowed for the eclass to modify the metadata of an ebuild, metadata being the
385 DEPENDS, RDEPENDS, SRC_URI, IUSE, etc, vars that are required to be constant,
386 and used by portage for dep resolution, fetching, etc. Using the earlier
387 example, if you're after a single function from an eclass (say epatch from
388 eutils), you -don't- want the metadata modifications the eclass you're
389 inheriting might do. You want to treat the eclass you're pulling from as a
390 library, pure and simple.</p>
391 <p>A new directory named elib should be added to the top level of the tree to serve
392 as a repository of ebuild function libraries. Rather then relying on using the
393 source command, an 'elib' function should be added to portage to import that
394 libraries functionality. The reason for the indirection via the function is
395 mostly related to portage internals, but it does serve as an abstraction such
396 that (for example) zsh compatibility hacks could be hidden in the elib function.</p>
397 <p>Elib's will be collections of bash functions- they're not allowed to do anything
398 in the global scope aside from function definition, and any -minimal-
399 initialization of the library that is absolutely needed. Additionally, they
400 cannot modify any ebuild template functions- src_compile, src_unpack. Since they are
401 required to not modify the metadata keys, nor in any way affect the ebuild aside
402 from providing functionality, they can be conditionally pulled in. They also
403 are allowed to pull in other elibs, but strictly just elibs- no eclasses, just
404 other elibs. A real world example would be the eutils eclass.</p>
405 <p>Portage, since the elib's don't modify metadata, isn't required to track elibs
406 as it tracks eclasses. Thus a change in an elib doesn't result in half the tree
407 forced to be regenerated/marked stale when changed (this is more of an infra
408 benefit, although regen's that take too long due to eclass changes have been
409 known to cause rsync issues due to missing timestamps).</p>
410 <p>Elibs will not be available in the global scope of an eclass, or ebuild- nor during the
411 depends phase (basically a phase that sources the ebuild, to get its metadata). Elib
412 calls in the global scope will be tracked, but the elib will not be loaded till just before
413 the setup phase (pkg_setup). There are two reasons for this- first, it ensures elibs are
414 completely incapable of modifying metadata. There is no room for confusion, late loading
415 of elibs gives you the functionality for all phases, except for depends- depends being the
416 only phase that is capable of specifying metadata. Second, as an added bonus, late
417 loading reduces the amount of bash sourced for a regen- faster regens. This however is minor,
418 and is an ancillary benefit of the first reason.</p>
419 <p>There are a few further restrictions with elibs--mainly, elibs to load can only be specified
420 in either global scope, or in the setup, unpack, compile, test, and install phases. You can
421 not load elibs in prerm, postrm, preinst, and postinst. The reason being, for *rm phases,
422 installed pkgs will have to look to the tree for the elib, which allows for api drift to cause
423 breakage. For *inst phases, same thing, except the culprit is binpkgs.</p>
424 <p>There is a final restriction--elibs cannot change their exported api dependent on the api
425 (as some eclass do for example). The reason mainly being that elibs are loaded once--not
426 multiple times, as eclasses are.</p>
427 <p>To clarify, for example this is invalid.</p>
428 <pre class="literal-block">
429 if [[ -n ${SOME_VAR} ]]; then
430 func x() { echo &quot;I'm accessible only via tweaking some var&quot;;}
431 else
432 func x() { echo &quot;this is invalid, do not do it.&quot;; }
433 fi
434 </pre>
435 <p>Regarding maintainability of elibs, it should be a less of a load then old
436 eclasses. One of the major issues with old eclasses is that their functions are
437 quite incestuous- they're bound tightly to the env they're defined in. This
438 makes eclass functions a bit fragile- the restrictions on what can, and cannot
439 be done in elibs will address this, making functionality less fragile (thus a
440 bit more maintainable).</p>
441 <p>There is no need for backwards compatibility with elibs- they just must work
442 against the current tree. Thus elibs can be removed when the tree no longer
443 needs them. The reasons for this are explained below.</p>
444 <p>Structuring of the elibs directory will be exactly the same as that of the new
445 eclass directory (detailed below), sans a different extension.</p>
446 <p>As to why their are so many restrictions, the answer is simple- the definition of
447 what elibs are, what they are capable of, and how to use them is nailed down as much as
448 possible to avoid <em>any</em> ambiguity related to them. The intention is to make it clear,
449 such that no misconceptions occur, resulting in bugs.</p>
450 </div>
451 <div class="section">
452 <h2><a class="toc-backref" href="#id8" id="the-reduced-role-of-eclasses-and-a-clarification-of-existing-eclass-requirements" name="the-reduced-role-of-eclasses-and-a-clarification-of-existing-eclass-requirements">The reduced role of Eclasses, and a clarification of existing Eclass requirements</a></h2>
453 <p>Since elibs are now intended on holding common bash functionality, the focus of
454 eclasses should be in defining an appropriate template for ebuilds. For example,
455 defining common DEPENDS, RDEPENDS, src_compile functions, src_unpack, etc.
456 Additionally, eclasses should pull in any elibs they need for functionality.</p>
457 <p>Eclass functionality that isn't directly related to the metadata, or src_* and
458 pkg_* funcs should be shifted into elibs to allow for maximal code reuse. This
459 however isn't a hard requirement, merely a strongly worded suggestion.</p>
460 <p>Previously, it was 'strongly' suggested by developers to avoid having any code
461 executed in the global scope that wasn't required. This suggestion is now a
462 requirement. Execute only what must be executed in the global scope. Any code
463 executed in the global scope that is related to configuring/building the package
464 must be placed in pkg_setup. Metadata keys (already a rule, but now stated as
465 an absolute requirement to clarify it) <em>must</em> be constant. The results of
466 metadata keys exported from an ebuild on system A, must be <em>exactly</em> the same as
467 the keys exported on system B.</p>
468 <p>If an eclass (or ebuild for that matter) violates this constant requirement, it
469 leads to portage doing the wrong thing for rsync users- for example, wrong deps
470 pulled in, leading to compilation failure, or dud deps.</p>
471 <p>If the existing metadata isn't flexible enough for what is required for a
472 package, the parsing of the metadata is changed to address that. Cases where
473 the constant requirement is violated are known, and a select few are allowed-
474 these are exceptions to the rule that are required due to inadequacies in
475 portage. Any case where it's determined the constant requirement may need to be
476 violated the dev must make it aware to the majority of devs, along with the portage
477 devs. This should be done prior to committing.</p>
478 <p>It's quite likely there is a way to allow what you're attempting- if you just go
479 and do it, the rsync users (our user base) suffer the results of compilation
480 failures and unneeded deps being pulled in.</p>
481 <p>After that stern reminder, back to new eclasses. Defining INHERITED and ECLASS
482 within the eclass is no longer required. Portage already handles those vars if
483 they aren't defined.</p>
484 <p>As with elibs, it's no longer required that backwards compatibility be maintained
485 indefinitely- compatibility must be maintained against the current tree, but
486 just that. As such new eclasses (the true distinction of new vs old is
487 elaborated in the next section) can be removed from the tree once they're no
488 longer in use.</p>
489 </div>
490 <div class="section">
491 <h2><a class="toc-backref" href="#id9" id="the-end-of-backwards-compatibility" name="the-end-of-backwards-compatibility">The end of backwards compatibility...</a></h2>
492 <p>With current eclasses, once the eclass is in use, its api can no longer be
493 changed, nor can the eclass ever be removed from the tree. This is why we still
494 have <em>ancient</em> eclasses that are completely unused sitting in the tree, for
495 example inherit.eclass. The reason for this, not surprisingly, is a portage
496 deficiency: on unmerging an installed ebuild, portage used the eclass from the
497 current tree.</p>
498 <p>For a real world example of this, if you merged a glibc 2 years back, whatever
499 eclasses it used must still be compatible, or you may not be able to unmerge the
500 older glibc version during an upgrade to a newer version. So either the glibc
501 maintainer is left with the option of leaving people using ancient versions out
502 in the rain, or maintaining an ever increasing load of backwards compatibility
503 cruft in any used eclasses.</p>
504 <p>Binpkgs suffer a similar fate. Merging of a binpkg pulls needed eclasses from
505 the tree, so you may not be able to even merge a binpkg if the eclasses api has
506 changed. If the eclass was removed, you can't even merge the binpkg, period.</p>
507 <p>The next major release of portage will address this- the environment that the
508 ebuild was built in already contains the eclasses functions, as such the env can
509 be re-used rather then relying on the eclass. In other words, binpkgs and
510 installed ebuilds will no longer go and pull needed eclasses from the tree,
511 they'll use the 'saved' version of the eclass they were built/merged with.</p>
512 <p>So the backwards compatibility requirement for users of the next major portage
513 version (and beyond) isn't required. All the cruft can be dropped.</p>
514 <p>The problem is that there will be users using older versions of portage that don't
515 support this functionality- these older installations <em>cannot</em> use the
516 new eclasses, due to the fact that their portage version is incapable of
517 properly relying on the env- in other words, the varying api of the eclass will
518 result in user-visible failures during unmerging.</p>
519 <p>So we're able to do a clean break of all old eclasses, and api cruft, but we need
520 a means to basically disallow access to the new eclasses for all portage versions
521 incapable of properly handling the env requirements.</p>
522 <p>Unfortunately, we cannot just rely on a different grouping/naming convention within
523 the old eclass directory. The new eclasses must be inaccessible, and portage throws
524 a snag into this- the existing inherit function that is used to handle existing
525 eclasses. Basically, whatever it's passed (inherit kernel or inherit
526 kernel/kernel) it will pull in (kernel.eclass, and kernel/kernel.eclass
527 respectively). So even if the new eclasses were implemented within a
528 subdirectory of the eclass dir in the tree, all current portage versions would
529 still be able to access them.</p>
530 <p>In other words, these new eclasses would in effect, be old eclasses since older
531 portage versions could still access them.</p>
532 </div>
533 <div class="section">
534 <h2><a class="toc-backref" href="#id10" id="tree-restructuring" name="tree-restructuring">Tree restructuring</a></h2>
535 <p>There are only two way to block the existing (as of this writing) inherit
536 functionality from accessing the new eclasses- either change the extension of
537 eclasses to something other then 'eclass', or to have them stored in a separate
538 subdirectory of the tree then eclass.</p>
539 <p>The latter is preferable, and the proposed solution. Reasons are- the current
540 eclass directory is already overgrown. Structuring of the new eclass dir
541 (clarified below) will allow for easier signing, ChangeLogs, and grouping of
542 eclasses. New eclasses allow for something akin to a clean break and have new
543 capabilities/requirements, thus it's advisable to start with a clean directory,
544 devoid of all cruft from the old eclass implementation.</p>
545 <p>If it's unclear as to why the old inherit function <em>cannot</em> access the new
546 eclasses, please reread the previous section. It's unfortunately a requirement
547 to take advantage of all that the next major portage release will allow.</p>
548 <p>The proposed directory structure is ${PORTDIR}/include/{eclass,elib}.
549 Something like ${PORTDIR}/new-eclass, or ${PORTDIR}/eclass-ng could be used
550 (although many would cringe at the -ng), but such a name is unwise. Consider the
551 possibility (likely a fact) that new eclasses someday may be found lacking, and
552 refined further (version three as it were). Or perhaps we want to add yet more
553 functionality with direct relation to sourcing new files, and we would then need
554 to further populate ${PORTDIR}.</p>
555 <p>The new-eclass directory will be (at least) 2 levels deep- for example:</p>
556 <dl class="docutils">
557 <dt>::</dt>
558 <dd>kernel/
559 kernel/linux-info.eclass
560 kernel/linux-mod.eclass
561 kernel/kernel-2.6.eclass
562 kernel/kernel-2.4.eclass
563 kernel/ChangeLog
564 kernel/Manifest</dd>
565 </dl>
566 <p>No eclasses will be allowed in the base directory- grouping of new eclasses will
567 be required to help keep things tidy, and for the following reasons. Grouping
568 of eclasses allows for the addition of ChangeLogs that are specific to that
569 group of eclasses, grouping of files/patches as needed, and allows for
570 saner/easier signing of eclasses- you can just stick a signed
571 Manifest file w/in that grouping, thus providing the information portage needs
572 to ensure no files are missing, and that nothing has been tainted.</p>
573 <p>The elib directory will be structured in the same way, for the same reasons.</p>
574 <p>Repoman will have to be extended to work within new eclass and elib groups, and
575 to handle signing and committing. This is intentional, and a good thing. This
576 gives repoman the possibility of doing sanity checks on elibs/new eclasses.</p>
577 <p>Note these checks will not prevent developers from doing dumb things with eclass-
578 these checks would only be capable of doing basic sanity checks, such as syntax checks.
579 There is no way to prevent people from doing dumb things (exempting perhaps repeated
580 applications of a cattle prod)- these are strictly automatic checks, akin to repoman's
581 dependency checks.</p>
582 </div>
583 <div class="section">
584 <h2><a class="toc-backref" href="#id11" id="the-start-of-a-different-phase-of-backwards-compatibility" name="the-start-of-a-different-phase-of-backwards-compatibility">The start of a different phase of backwards compatibility</a></h2>
585 <p>As clarified above, new eclasses will exist in a separate directory that will be
586 intentionally inaccessible to the inherit function. As such, users of older
587 portage versions <em>will</em> have to upgrade to merge any ebuild that uses elibs/new
588 eclasses. A depend on the next major portage version would transparently handle
589 this for rsync users.</p>
590 <p>There still is the issue of users who haven't upgraded to the required portage
591 version. This is a minor concern frankly- portage releases include new
592 functionality, and bug fixes. If they won't upgrade, it's assumed they have
593 their reasons and are big boys, thus able to handle the complications themselves.</p>
594 <p>The real issue is broken envs, whether in binpkgs, or for installed packages.
595 Two options exist- either the old eclasses are left in the tree indefinitely, or
596 they're left for N months, then shifted out of the tree, and into a tarball that
597 can be merged.</p>
598 <p>Shifting them out of the tree is advisable for several reasons- less cruft in
599 the tree, but more importantly the fact that they are not signed (thus an angle
600 for attack). Note that the proposed method of eclass signing doesn't even try
601 to address them. Frankly, it's not worth the effort supporting two variations
602 of eclass signing, when the old eclass setup isn't designed to allow for easy
603 signing.</p>
604 <p>If this approach is taken, then either the old eclasses would have to be merged
605 to an overlay directory's eclass directory (ugly), or to a safe location that
606 portage's inherit function knows to look for (less ugly).</p>
607 <p>For users who do not upgrade within the window of N months while the old
608 eclasses are in the tree, as stated, it's assumed they know what they are doing.
609 If they specifically block the new portage version, as the ebuilds in the tree
610 migrate to the new eclasses, they will have less and less ebuilds available to
611 them. If they tried injecting the new portage version (lying to portage,
612 essentially), portage would bail since it cannot find the new eclass.
613 For ebuilds that use the new eclasses, there really isn't any way to sidestep
614 the portage version requirement- same as it has been for other portage features.</p>
615 <p>What is a bit more annoying is that once the old eclasses are out of the tree,
616 if a user has not upgraded to a portage version supporting env processing, they
617 will lose the ability to unmerge any installed ebuild that used an old
618 eclass. Same cause, different symptom being they will lose the ability to merge
619 any tbz2 that uses old eclasses also.</p>
620 <p>There is one additional case that is a rarity, but should be noted- if a user
621 has suffered significant corruption of their installed package database (vdb). This is
622 ignoring the question of whether the vdb is even usable at this point, but the possibility
623 exists for the saved envs to be non usable due to either A) missing, or B) corrupted.
624 In such a case, even with the new portage capabilities, they would need
625 the old eclass compat ebuild.</p>
626 <p>Note for this to happen requires either rather... unwise uses of root, or significant
627 fs corruption. Regardless of the cause, it's quite likely for this to even become an
628 issue, the system's vdb is completely unusable. It's a moot issue at that point.
629 If you lose your vdb, or it gets seriously damaged, it's akin to lobotomizing portage-
630 it doesn't know what's installed, it doesn't know of its own files, and in general,
631 a rebuilding of the system is about the only sane course of action. The missing env is
632 truly the least of the users concern in such a case.</p>
633 <p>Continuing with the more likely scenario, users unwilling to upgrade portage will
634 <em>not</em> be left out in the rain. Merging the old eclass compat ebuild will provide
635 the missing eclasses, thus providing that lost functionality.</p>
636 <p>Note the intention isn't to force them to upgrade, hence the ability to restore the
637 lost functionality. The intention is to clean up the existing mess, and allow us
638 to move forward. The saying &quot;you've got to break a few eggs to make an omelet&quot;
639 is akin, exempting the fact we're providing a way to make the eggs whole again
640 (the king's men would've loved such an option).</p>
641 </div>
642 <div class="section">
643 <h2><a class="toc-backref" href="#id12" id="migrating-to-the-new-setup" name="migrating-to-the-new-setup">Migrating to the new setup</a></h2>
644 <p>As has been done in the past whenever a change in the tree results in ebuilds
645 requiring a specific version of portage, as ebuilds migrate to the new eclasses,
646 they should depend on a version of portage that supports it. From the users
647 viewpoint, this transparently handles the migration.</p>
648 <p>This isn't so transparent for devs or a particular infrastructure server however.
649 Devs, due to them using cvs for their tree, lack the pregenerated cache rsync
650 users have. Devs will have to be early adopters of the new portage. Older
651 portage versions won't be able to access the new eclasses, thus the local cache
652 generation for that ebuild will fail, ergo the depends on a newer portage
653 version won't transparently handle it for them.</p>
654 <p>Additionally, prior to any ebuilds in the tree using the new eclasses, the
655 infrastructure server that generates the cache for rsync users will have to
656 either be upgraded to a version of portage supporting new eclasses, or patched.
657 The former being much more preferable then the latter for the portage devs.</p>
658 <p>Beyond that, an appropriate window for old eclasses to exist in the tree must be
659 determined, and prior to that window passing, an ebuild must be added to the tree
660 so users can get the old eclasses if needed.</p>
661 <p>For eclass devs to migrate from old to new, it is possible for them to just
662 transfer the old eclass into an appropriate grouping in the new eclass directory,
663 although it's advisable they cleanse all cruft out of the eclass. You can
664 migrate ebuilds gradually over to the new eclass, and don't have to worry about
665 having to support ebuilds from X years back.</p>
666 <p>Essentially, you have a chance to nail the design perfectly/cleanly, and have a
667 window in which to redesign it. It's humbly suggested eclass devs take
668 advantage of it. :)</p>
669 </div>
670 </div>
671 <div class="section">
672 <h1><a class="toc-backref" href="#id13" id="backwards-compatibility" name="backwards-compatibility">Backwards Compatibility</a></h1>
673 <p>All backwards compatibility issues are addressed in line, but a recap is offered-
674 it's suggested that if the a particular compatibility issue is
675 questioned/worried over, the reader read the relevant section. There should be
676 a more in depth discussion of the issue, along with a more extensive explanation
677 of the potential solutions, and reasons for the chosen solution.</p>
678 <p>To recap:</p>
679 <pre class="literal-block">
680 New eclasses and elib functionality will be tied to a specific portage
681 version. A DEPENDs on said portage version should address this for rsync
682 users who refuse to upgrade to a portage version that supports the new
683 eclasses/elibs and will gradually be unable to merge ebuilds that use said
684 functionality. It is their choice to upgrade, as such, the gradual
685 'thinning' of available ebuilds should they block the portage upgrade is
686 their responsibility.
688 Old eclasses at some point in the future should be removed from the tree,
689 and released in a tarball/ebuild. This will cause installed ebuilds that
690 rely on the old eclass to be unable to unmerge, with the same applying for
691 merging of binpkgs dependent on the following paragraph.
693 The old eclass-compat is only required for users who do not upgrade their
694 portage installation, and one further exemption- if the user has somehow
695 corrupted/destroyed their installed pkgs database (/var/db/pkg currently),
696 in the process, they've lost their saved environments. The eclass-compat
697 ebuild would be required for ebuilds that required older eclasses in such a
698 case. Note, this case is rare also- as clarified above, it's mentioned
699 strictly to be complete, it's not much of a real world scenario as elaborated
700 above.
701 </pre>
702 </div>
703 <div class="section">
704 <h1><a class="toc-backref" href="#id14" id="copyright" name="copyright">Copyright</a></h1>
705 <p>This document has been placed in the public domain.</p>
706 </div>
708 </div>
709 <div class="footer">
710 <hr class="footer" />
711 <a class="reference" href="glep-0033.txt">View document source</a>.
712 Generated on: 2006-10-10 20:23 UTC.
713 Generated by <a class="reference" href="http://docutils.sourceforge.net/">Docutils</a> from <a class="reference" href="http://docutils.sourceforge.net/rst.html">reStructuredText</a> source.
715 </div>
716 </body>
717 </html>

  ViewVC Help
Powered by ViewVC 1.1.20