/[gentoo]/xml/htdocs/proj/en/glep/glep-0033.html
Gentoo

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

Parent Directory Parent Directory | Revision Log Revision Log


Revision 1.3 - (hide annotations) (download) (as text)
Thu Sep 15 02:37:38 2005 UTC (8 years, 10 months ago) by vapier
Branch: MAIN
Changes since 1.2: +40 -40 lines
File MIME type: text/html
clean up grammer and punctuation

1 g2boojum 1.1 <?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
7     to templates. DO NOT USE THIS HTML FILE AS YOUR TEMPLATE!
8     -->
9     <head>
10     <meta http-equiv="Content-Type" content="text/html; charset=utf-8" />
11 vapier 1.3 <meta name="generator" content="Docutils 0.3.9: http://docutils.sourceforge.net/" />
12 g2boojum 1.1 <title>GLEP 33 -- Eclass Restructure/Redesign</title>
13     <link rel="stylesheet" href="tools/glep.css" type="text/css" />
14     </head>
15     <body bgcolor="white">
16     <table class="navigation" cellpadding="0" cellspacing="0"
17     width="100%" border="0">
18     <tr><td class="navicon" width="150" height="35">
19     <a href="http://www.gentoo.org/" title="Gentoo Linux Home Page">
20     <img src="http://www.gentoo.org/images/gentoo-new.gif" alt="[Gentoo]"
21     border="0" width="150" height="35" /></a></td>
22     <td class="textlinks" align="left">
23     [<b><a href="http://www.gentoo.org/">Gentoo Linux Home</a></b>]
24     [<b><a href="http://www.gentoo.org/proj/en/glep">GLEP Index</a></b>]
25     [<b><a href="./glep-0033.txt">GLEP Source</a></b>]
26     </td></tr></table>
27 vapier 1.3 <table class="rfc2822 docutils field-list" frame="void" rules="none">
28 g2boojum 1.1 <col class="field-name" />
29     <col class="field-body" />
30     <tbody valign="top">
31     <tr class="field"><th class="field-name">GLEP:</th><td class="field-body">33</td>
32     </tr>
33     <tr class="field"><th class="field-name">Title:</th><td class="field-body">Eclass Restructure/Redesign</td>
34     </tr>
35 vapier 1.3 <tr class="field"><th class="field-name">Version:</th><td class="field-body">1.3</td>
36 g2boojum 1.1 </tr>
37 vapier 1.3 <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/xml/htdocs/proj/en/glep/glep-0033.txt?cvsroot=gentoo">2005/03/06 20:39:19</a></td>
38 g2boojum 1.1 </tr>
39 g2boojum 1.2 <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>
40 g2boojum 1.1 </tr>
41     <tr class="field"><th class="field-name">Status:</th><td class="field-body">Draft</td>
42     </tr>
43     <tr class="field"><th class="field-name">Type:</th><td class="field-body">Standards Track</td>
44     </tr>
45 vapier 1.3 <tr class="field"><th class="field-name">Content-Type:</th><td class="field-body"><a class="reference" href="http://www.python.org/peps/glep-0012.html">text/x-rst</a></td>
46 g2boojum 1.1 </tr>
47     <tr class="field"><th class="field-name">Created:</th><td class="field-body">29-Jan-2005</td>
48     </tr>
49 vapier 1.3 <tr class="field"><th class="field-name">Post-History:</th><td class="field-body">29-Jan-2005 6-Mar-2005</td>
50 g2boojum 1.1 </tr>
51     </tbody>
52     </table>
53     <hr />
54     <div class="contents topic" id="contents">
55     <p class="topic-title first"><a name="contents">Contents</a></p>
56     <ul class="simple">
57     <li><a class="reference" href="#abstract" id="id2" name="id2">Abstract</a></li>
58     <li><a class="reference" href="#terminology" id="id3" name="id3">Terminology</a></li>
59     <li><a class="reference" href="#motivation-and-rationale" id="id4" name="id4">Motivation and Rationale</a></li>
60 vapier 1.3 <li><a class="reference" href="#specification" id="id5" name="id5">Specification</a><ul>
61 g2boojum 1.1 <li><a class="reference" href="#ebuild-libraries-elibs-for-short" id="id6" name="id6">Ebuild Libraries (elibs for short)</a></li>
62     <li><a class="reference" href="#the-reduced-role-of-eclasses-and-a-clarification-of-existing-eclass-requirements" id="id7" name="id7">The reduced role of Eclasses, and a clarification of existing Eclass requirements</a></li>
63 g2boojum 1.2 <li><a class="reference" href="#the-end-of-backwards-compatibility" id="id8" name="id8">The end of backwards compatibility...</a></li>
64 vapier 1.3 <li><a class="reference" href="#tree-restructuring" id="id9" name="id9">Tree restructuring</a></li>
65 g2boojum 1.2 <li><a class="reference" href="#the-start-of-a-different-phase-of-backwards-compatibility" id="id10" name="id10">The start of a different phase of backwards compatibility</a></li>
66 g2boojum 1.1 <li><a class="reference" href="#migrating-to-the-new-setup" id="id11" name="id11">Migrating to the new setup</a></li>
67     </ul>
68     </li>
69     <li><a class="reference" href="#backwards-compatibility" id="id12" name="id12">Backwards Compatibility</a></li>
70     <li><a class="reference" href="#copyright" id="id13" name="id13">Copyright</a></li>
71     </ul>
72     </div>
73     <div class="section" id="abstract">
74     <h1><a class="toc-backref" href="#id2" name="abstract">Abstract</a></h1>
75     <p>For any design, the transition from theoretical to applied exposes inadequacies
76     in the original design. This document is intended to document, and propose a
77     revision of the current eclass setup to address current eclass inadequacies.</p>
78 g2boojum 1.2 <p>This document proposes several things- the creation of ebuild libraries, 'elibs',
79 g2boojum 1.1 a narrowing of the focus of eclasses, a move of eclasses w/in the tree, the
80     addition of changelogs, and a way to allow for simple eclass gpg signing.
81     In general, a large scale restructuring of what eclasses are and how they're
82     implemented. Essentially version two of the eclass setup.</p>
83     </div>
84     <div class="section" id="terminology">
85     <h1><a class="toc-backref" href="#id3" name="terminology">Terminology</a></h1>
86     <p>From this point on, the proposed eclass setup will be called 'new eclasses', the
87     existing crop (as of this writing) will be referenced as 'old eclasses'. The
88 g2boojum 1.2 distinction is elaborated on within this document.</p>
89 g2boojum 1.1 </div>
90     <div class="section" id="motivation-and-rationale">
91     <h1><a class="toc-backref" href="#id4" name="motivation-and-rationale">Motivation and Rationale</a></h1>
92     <p>Eclasses within the tree currently are a bit of a mess- they're forced to
93 g2boojum 1.2 maintain backwards compatibility w/ all previous functionality. In effect,
94 g2boojum 1.1 their api is constant, and can only be added to- never changing the existing
95 vapier 1.3 functionality. This obviously is quite limiting, and leads to cruft accruing in
96 g2boojum 1.1 eclasses as a eclasses design is refined. This needs to be dealt with prior to
97 g2boojum 1.2 eclass code reaching a critical mass where they become unmanageable/fragile
98     (recent pushes for eclass versioning could be interpreted as proof of this).</p>
99 g2boojum 1.1 <p>Beyond that, eclasses were originally intended as a method to allow for ebuilds
100     to use a pre-existing block of code, rather then having to duplicate the code in
101     each ebuild. This is a good thing, but there are ill effects that result from
102 vapier 1.3 the current design. Eclasses inherit other eclasses to get a single function- in
103 g2boojum 1.1 doing so, modifying the the exported 'template' (default src_compile, default
104 vapier 1.3 src_unpack, various vars, etc). All the eclass designer was after was reusing a
105 g2boojum 1.1 function, not making their eclass sensitive to changes in the template of the
106     eclass it's inheriting. The eclass designer -should- be aware of changes in the
107     function they're using, but shouldn't have to worry about their default src_*
108     and pkg_* functions being overwritten, let alone the env changes.</p>
109     <p>Addressing up front why a collection of eclass refinements are being rolled into
110     a single set of changes, parts of this proposal -could- be split into multiple
111     phases. Why do it though? It's simpler for developers to know that the first
112     eclass specification was this, and that the second specification is that,
113     rather then requiring them to be aware of what phase of eclass changes is in
114     progress.</p>
115     <p>By rolling all changes into one large change, a line is intentionally drawn in
116     the sand. Old eclasses allowed for this, behaved this way. New eclasses allow
117 vapier 1.3 for that, and behave this way. This should reduce misconceptions about what is
118 g2boojum 1.1 allowed/possible with eclasses, thus reducing bugs that result from said
119     misconceptions.</p>
120 g2boojum 1.2 <p>A few words on elibs- think of them as a clear definition between behavioral
121     functionality of an eclass, and the library functionality. Eclass's modify
122     template data, and are the basis for other ebuilds- elibs, however are <em>just</em>
123     common bash functionality.</p>
124     <p>Consider the majority of the portage bin/* scripts- these all are candidates for
125     being added to the tree as elibs, as is the bulk of eutils.</p>
126 g2boojum 1.1 </div>
127     <div class="section" id="specification">
128 vapier 1.3 <h1><a class="toc-backref" href="#id5" name="specification">Specification</a></h1>
129 g2boojum 1.1 <p>The various parts of this proposal are broken down into a set of changes and
130     elaborations on why a proposed change is preferable. It's advisable to the
131     reader that this be read serially, rather then jumping around.</p>
132     <div class="section" id="ebuild-libraries-elibs-for-short">
133     <h2><a class="toc-backref" href="#id6" name="ebuild-libraries-elibs-for-short">Ebuild Libraries (elibs for short)</a></h2>
134     <p>As briefly touched upon in Motivation and Rationale, the original eclass design
135     allowed for the eclass to modify the metadata of an ebuild, metadata being the
136     DEPENDS, RDEPENDS, SRC_URI, IUSE, etc, vars that are required to be constant,
137     and used by portage for dep resolution, fetching, etc. Using the earlier
138     example, if you're after a single function from an eclass (say epatch from
139     eutils), you -don't- want the metadata modifications the eclass you're
140     inheriting might do. You want to treat the eclass you're pulling from as a
141     library, pure and simple.</p>
142     <p>A new directory named elib should be added to the top level of the tree to serve
143     as a repository of ebuild function libraries. Rather then relying on using the
144     source command, an 'elib' function should be added to portage to import that
145     libraries functionality. The reason for the indirection via the function is
146     mostly related to portage internals, but it does serve as an abstraction such
147 g2boojum 1.2 that (for example) zsh compatibility hacks could be hidden in the elib function.</p>
148 g2boojum 1.1 <p>Elib's will be collections of bash functions- they're not allowed to do anything
149     in the global scope aside from function definition, and any -minimal-
150     initialization of the library that is absolutely needed. Additionally, they
151 g2boojum 1.2 cannot modify any ebuild template functions- src_compile, src_unpack. Since they are
152 g2boojum 1.1 required to not modify the metadata keys, nor in any way affect the ebuild aside
153     from providing functionality, they can be conditionally pulled in. They also
154     are allowed to pull in other elibs, but strictly just elibs- no eclasses, just
155 vapier 1.3 other elibs. A real world example would be the eutils eclass.</p>
156 g2boojum 1.1 <p>Portage, since the elib's don't modify metadata, isn't required to track elibs
157     as it tracks eclasses. Thus a change in an elib doesn't result in half the tree
158     forced to be regenerated/marked stale when changed (this is more of an infra
159     benefit, although regen's that take too long due to eclass changes have been
160 g2boojum 1.2 known to cause rsync issues due to missing timestamps).</p>
161     <p>Elibs will not be available in the global scope of an eclass, or ebuild- nor during the
162 vapier 1.3 depends phase (basically a phase that sources the ebuild, to get its metadata). Elib
163 g2boojum 1.2 calls in the global scope will be tracked, but the elib will not be loaded till just before
164 vapier 1.3 the setup phase (pkg_setup). There are two reasons for this- first, it ensures elibs are
165 g2boojum 1.2 completely incapable of modifying metadata. There is no room for confusion, late loading
166     of elibs gives you the functionality for all phases, except for depends- depends being the
167     only phase that is capable of specifying metadata. Second, as an added bonus, late
168     loading reduces the amount of bash sourced for a regen- faster regens. This however is minor,
169     and is an ancillary benefit of the first reason.</p>
170     <p>There are a few further restrictions with elibs--mainly, elibs to load can only be specified
171     in either global scope, or in the setup, unpack, compile, test, and install phases. You can
172     not load elibs in prerm, postrm, preinst, and postinst. The reason being, for *rm phases,
173     installed pkgs will have to look to the tree for the elib, which allows for api drift to cause
174     breakage. For *inst phases, same thing, except the culprit is binpkgs.</p>
175     <p>There is a final restriction--elibs cannot change their exported api dependent on the api
176     (as some eclass do for example). The reason mainly being that elibs are loaded once--not
177     multiple times, as eclasses are.</p>
178     <p>To clarify, for example this is invalid.</p>
179     <pre class="literal-block">
180     if [[ -n ${SOME_VAR} ]]; then
181     func x() { echo &quot;I'm accessible only via tweaking some var&quot;;}
182     else
183     func x() { echo &quot;this is invalid, do not do it.&quot;; }
184     fi
185     </pre>
186 g2boojum 1.1 <p>Regarding maintainability of elibs, it should be a less of a load then old
187     eclasses. One of the major issues with old eclasses is that their functions are
188 vapier 1.3 quite incestuous- they're bound tightly to the env they're defined in. This
189 g2boojum 1.1 makes eclass functions a bit fragile- the restrictions on what can, and cannot
190     be done in elibs will address this, making functionality less fragile (thus a
191     bit more maintainable).</p>
192 g2boojum 1.2 <p>There is no need for backwards compatibility with elibs- they just must work
193 g2boojum 1.1 against the current tree. Thus elibs can be removed when the tree no longer
194     needs them. The reasons for this are explained below.</p>
195     <p>Structuring of the elibs directory will be exactly the same as that of the new
196     eclass directory (detailed below), sans a different extension.</p>
197 g2boojum 1.2 <p>As to why their are so many restrictions, the answer is simple- the definition of
198     what elibs are, what they are capable of, and how to use them is nailed down as much as
199     possible to avoid <em>any</em> ambiguity related to them. The intention is to make it clear,
200     such that no misconceptions occur, resulting in bugs.</p>
201 g2boojum 1.1 </div>
202     <div class="section" id="the-reduced-role-of-eclasses-and-a-clarification-of-existing-eclass-requirements">
203     <h2><a class="toc-backref" href="#id7" 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>
204     <p>Since elibs are now intended on holding common bash functionality, the focus of
205 vapier 1.3 eclasses should be in defining an appropriate template for ebuilds. For example,
206 g2boojum 1.1 defining common DEPENDS, RDEPENDS, src_compile functions, src_unpack, etc.
207     Additionally, eclasses should pull in any elibs they need for functionality.</p>
208     <p>Eclass functionality that isn't directly related to the metadata, or src_* and
209     pkg_* funcs should be shifted into elibs to allow for maximal code reuse. This
210     however isn't a hard requirement, merely a strongly worded suggestion.</p>
211     <p>Previously, it was 'strongly' suggested by developers to avoid having any code
212     executed in the global scope that wasn't required. This suggestion is now a
213     requirement. Execute only what must be executed in the global scope. Any code
214     executed in the global scope that is related to configuring/building the package
215     must be placed in pkg_setup. Metadata keys (already a rule, but now stated as
216     an absolute requirement to clarify it) <em>must</em> be constant. The results of
217     metadata keys exported from an ebuild on system A, must be <em>exactly</em> the same as
218     the keys exported on system B.</p>
219     <p>If an eclass (or ebuild for that matter) violates this constant requirement, it
220     leads to portage doing the wrong thing for rsync users- for example, wrong deps
221 g2boojum 1.2 pulled in, leading to compilation failure, or dud deps.</p>
222 g2boojum 1.1 <p>If the existing metadata isn't flexible enough for what is required for a
223     package, the parsing of the metadata is changed to address that. Cases where
224     the constant requirement is violated are known, and a select few are allowed-
225     these are exceptions to the rule that are required due to inadequacies in
226 g2boojum 1.2 portage. Any case where it's determined the constant requirement may need to be
227     violated the dev must make it aware to the majority of devs, along with the portage
228     devs. This should be done prior to committing.</p>
229 g2boojum 1.1 <p>It's quite likely there is a way to allow what you're attempting- if you just go
230 g2boojum 1.2 and do it, the rsync users (our user base) suffer the results of compilation
231 g2boojum 1.1 failures and unneeded deps being pulled in.</p>
232     <p>After that stern reminder, back to new eclasses. Defining INHERITED and ECLASS
233     within the eclass is no longer required. Portage already handles those vars if
234     they aren't defined.</p>
235 g2boojum 1.2 <p>As with elibs, it's no longer required that backwards compatibility be maintained
236     indefinitely- compatibility must be maintained against the current tree, but
237 g2boojum 1.1 just that. As such new eclasses (the true distinction of new vs old is
238     elaborated in the next section) can be removed from the tree once they're no
239     longer in use.</p>
240     </div>
241 g2boojum 1.2 <div class="section" id="the-end-of-backwards-compatibility">
242     <h2><a class="toc-backref" href="#id8" name="the-end-of-backwards-compatibility">The end of backwards compatibility...</a></h2>
243 vapier 1.3 <p>With current eclasses, once the eclass is in use, its api can no longer be
244 g2boojum 1.1 changed, nor can the eclass ever be removed from the tree. This is why we still
245     have <em>ancient</em> eclasses that are completely unused sitting in the tree, for
246 vapier 1.3 example inherit.eclass. The reason for this, not surprisingly, is a portage
247     deficiency: on unmerging an installed ebuild, portage used the eclass from the
248 g2boojum 1.1 current tree.</p>
249     <p>For a real world example of this, if you merged a glibc 2 years back, whatever
250     eclasses it used must still be compatible, or you may not be able to unmerge the
251     older glibc version during an upgrade to a newer version. So either the glibc
252     maintainer is left with the option of leaving people using ancient versions out
253 g2boojum 1.2 in the rain, or maintaining an ever increasing load of backwards compatibility
254 g2boojum 1.1 cruft in any used eclasses.</p>
255     <p>Binpkgs suffer a similar fate. Merging of a binpkg pulls needed eclasses from
256     the tree, so you may not be able to even merge a binpkg if the eclasses api has
257     changed. If the eclass was removed, you can't even merge the binpkg, period.</p>
258     <p>The next major release of portage will address this- the environment that the
259     ebuild was built in already contains the eclasses functions, as such the env can
260     be re-used rather then relying on the eclass. In other words, binpkgs and
261     installed ebuilds will no longer go and pull needed eclasses from the tree,
262     they'll use the 'saved' version of the eclass they were built/merged with.</p>
263 g2boojum 1.2 <p>So the backwards compatibility requirement for users of the next major portage
264 g2boojum 1.1 version (and beyond) isn't required. All the cruft can be dropped.</p>
265 g2boojum 1.2 <p>The problem is that there will be users using older versions of portage that don't
266     support this functionality- these older installations <em>cannot</em> use the
267     new eclasses, due to the fact that their portage version is incapable of
268     properly relying on the env- in other words, the varying api of the eclass will
269     result in user-visible failures during unmerging.</p>
270     <p>So we're able to do a clean break of all old eclasses, and api cruft, but we need
271     a means to basically disallow access to the new eclasses for all portage versions
272     incapable of properly handling the env requirements.</p>
273     <p>Unfortunately, we cannot just rely on a different grouping/naming convention within
274     the old eclass directory. The new eclasses must be inaccessible, and portage throws
275     a snag into this- the existing inherit function that is used to handle existing
276     eclasses. Basically, whatever it's passed (inherit kernel or inherit
277 g2boojum 1.1 kernel/kernel) it will pull in (kernel.eclass, and kernel/kernel.eclass
278     respectively). So even if the new eclasses were implemented within a
279     subdirectory of the eclass dir in the tree, all current portage versions would
280     still be able to access them.</p>
281     <p>In other words, these new eclasses would in effect, be old eclasses since older
282     portage versions could still access them.</p>
283     </div>
284     <div class="section" id="tree-restructuring">
285 vapier 1.3 <h2><a class="toc-backref" href="#id9" name="tree-restructuring">Tree restructuring</a></h2>
286 g2boojum 1.1 <p>There are only two way to block the existing (as of this writing) inherit
287     functionality from accessing the new eclasses- either change the extension of
288 g2boojum 1.2 eclasses to something other then 'eclass', or to have them stored in a separate
289 g2boojum 1.1 subdirectory of the tree then eclass.</p>
290     <p>The latter is preferable, and the proposed solution. Reasons are- the current
291     eclass directory is already overgrown. Structuring of the new eclass dir
292     (clarified below) will allow for easier signing, ChangeLogs, and grouping of
293     eclasses. New eclasses allow for something akin to a clean break and have new
294     capabilities/requirements, thus it's advisable to start with a clean directory,
295     devoid of all cruft from the old eclass implementation.</p>
296     <p>If it's unclear as to why the old inherit function <em>cannot</em> access the new
297     eclasses, please reread the previous section. It's unfortunately a requirement
298     to take advantage of all that the next major portage release will allow.</p>
299 g2boojum 1.2 <p>The proposed directory structure is ${PORTDIR}/include/{eclass,elib}.
300 g2boojum 1.1 Something like ${PORTDIR}/new-eclass, or ${PORTDIR}/eclass-ng could be used
301 vapier 1.3 (although many would cringe at the -ng), but such a name is unwise. Consider the
302 g2boojum 1.1 possibility (likely a fact) that new eclasses someday may be found lacking, and
303 vapier 1.3 refined further (version three as it were). Or perhaps we want to add yet more
304 g2boojum 1.1 functionality with direct relation to sourcing new files, and we would then need
305     to further populate ${PORTDIR}.</p>
306     <p>The new-eclass directory will be (at least) 2 levels deep- for example:</p>
307 vapier 1.3 <dl class="docutils">
308 g2boojum 1.1 <dt>::</dt>
309     <dd>kernel/
310     kernel/linux-info.eclass
311     kernel/linux-mod.eclass
312     kernel/kernel-2.6.eclass
313     kernel/kernel-2.4.eclass
314     kernel/ChangeLog
315     kernel/Manifest</dd>
316     </dl>
317     <p>No eclasses will be allowed in the base directory- grouping of new eclasses will
318     be required to help keep things tidy, and for the following reasons. Grouping
319     of eclasses allows for the addition of ChangeLogs that are specific to that
320     group of eclasses, grouping of files/patches as needed, and allows for
321 g2boojum 1.2 saner/easier signing of eclasses- you can just stick a signed
322 g2boojum 1.1 Manifest file w/in that grouping, thus providing the information portage needs
323     to ensure no files are missing, and that nothing has been tainted.</p>
324     <p>The elib directory will be structured in the same way, for the same reasons.</p>
325     <p>Repoman will have to be extended to work within new eclass and elib groups, and
326 vapier 1.3 to handle signing and committing. This is intentional, and a good thing. This
327 g2boojum 1.2 gives repoman the possibility of doing sanity checks on elibs/new eclasses.</p>
328     <p>Note these checks will not prevent developers from doing dumb things with eclass-
329     these checks would only be capable of doing basic sanity checks, such as syntax checks.
330     There is no way to prevent people from doing dumb things (exempting perhaps repeated
331     applications of a cattle prod)- these are strictly automatic checks, akin to repoman's
332     dependency checks.</p>
333     </div>
334     <div class="section" id="the-start-of-a-different-phase-of-backwards-compatibility">
335     <h2><a class="toc-backref" href="#id10" name="the-start-of-a-different-phase-of-backwards-compatibility">The start of a different phase of backwards compatibility</a></h2>
336     <p>As clarified above, new eclasses will exist in a separate directory that will be
337 g2boojum 1.1 intentionally inaccessible to the inherit function. As such, users of older
338     portage versions <em>will</em> have to upgrade to merge any ebuild that uses elibs/new
339 vapier 1.3 eclasses. A depend on the next major portage version would transparently handle
340 g2boojum 1.2 this for rsync users.</p>
341 g2boojum 1.1 <p>There still is the issue of users who haven't upgraded to the required portage
342 vapier 1.3 version. This is a minor concern frankly- portage releases include new
343 g2boojum 1.1 functionality, and bug fixes. If they won't upgrade, it's assumed they have
344     their reasons and are big boys, thus able to handle the complications themselves.</p>
345     <p>The real issue is broken envs, whether in binpkgs, or for installed packages.
346     Two options exist- either the old eclasses are left in the tree indefinitely, or
347     they're left for N months, then shifted out of the tree, and into a tarball that
348     can be merged.</p>
349     <p>Shifting them out of the tree is advisable for several reasons- less cruft in
350     the tree, but more importantly the fact that they are not signed (thus an angle
351     for attack). Note that the proposed method of eclass signing doesn't even try
352     to address them. Frankly, it's not worth the effort supporting two variations
353     of eclass signing, when the old eclass setup isn't designed to allow for easy
354     signing.</p>
355     <p>If this approach is taken, then either the old eclasses would have to be merged
356     to an overlay directory's eclass directory (ugly), or to a safe location that
357     portage's inherit function knows to look for (less ugly).</p>
358     <p>For users who do not upgrade within the window of N months while the old
359     eclasses are in the tree, as stated, it's assumed they know what they are doing.
360     If they specifically block the new portage version, as the ebuilds in the tree
361     migrate to the new eclasses, they will have less and less ebuilds available to
362 g2boojum 1.2 them. If they tried injecting the new portage version (lying to portage,
363     essentially), portage would bail since it cannot find the new eclass.
364     For ebuilds that use the new eclasses, there really isn't any way to sidestep
365     the portage version requirement- same as it has been for other portage features.</p>
366 g2boojum 1.1 <p>What is a bit more annoying is that once the old eclasses are out of the tree,
367 g2boojum 1.2 if a user has not upgraded to a portage version supporting env processing, they
368     will lose the ability to unmerge any installed ebuild that used an old
369     eclass. Same cause, different symptom being they will lose the ability to merge
370     any tbz2 that uses old eclasses also.</p>
371     <p>There is one additional case that is a rarity, but should be noted- if a user
372     has suffered significant corruption of their installed package database (vdb). This is
373     ignoring the question of whether the vdb is even usable at this point, but the possibility
374     exists for the saved envs to be non usable due to either A) missing, or B) corrupted.
375     In such a case, even with the new portage capabilities, they would need
376     the old eclass compat ebuild.</p>
377     <p>Note for this to happen requires either rather... unwise uses of root, or significant
378     fs corruption. Regardless of the cause, it's quite likely for this to even become an
379     issue, the system's vdb is completely unusable. It's a moot issue at that point.
380     If you lose your vdb, or it gets seriously damaged, it's akin to lobotomizing portage-
381 vapier 1.3 it doesn't know what's installed, it doesn't know of its own files, and in general,
382 g2boojum 1.2 a rebuilding of the system is about the only sane course of action. The missing env is
383     truly the least of the users concern in such a case.</p>
384     <p>Continuing with the more likely scenario, users unwilling to upgrade portage will
385     <em>not</em> be left out in the rain. Merging the old eclass compat ebuild will provide
386 vapier 1.3 the missing eclasses, thus providing that lost functionality.</p>
387 g2boojum 1.2 <p>Note the intention isn't to force them to upgrade, hence the ability to restore the
388 vapier 1.3 lost functionality. The intention is to clean up the existing mess, and allow us
389     to move forward. The saying &quot;you've got to break a few eggs to make an omelet&quot;
390 g2boojum 1.1 is akin, exempting the fact we're providing a way to make the eggs whole again
391     (the king's men would've loved such an option).</p>
392     </div>
393     <div class="section" id="migrating-to-the-new-setup">
394     <h2><a class="toc-backref" href="#id11" name="migrating-to-the-new-setup">Migrating to the new setup</a></h2>
395     <p>As has been done in the past whenever a change in the tree results in ebuilds
396     requiring a specific version of portage, as ebuilds migrate to the new eclasses,
397     they should depend on a version of portage that supports it. From the users
398     viewpoint, this transparently handles the migration.</p>
399     <p>This isn't so transparent for devs or a particular infrastructure server however.
400     Devs, due to them using cvs for their tree, lack the pregenerated cache rsync
401     users have. Devs will have to be early adopters of the new portage. Older
402     portage versions won't be able to access the new eclasses, thus the local cache
403     generation for that ebuild will fail, ergo the depends on a newer portage
404     version won't transparently handle it for them.</p>
405     <p>Additionally, prior to any ebuilds in the tree using the new eclasses, the
406     infrastructure server that generates the cache for rsync users will have to
407     either be upgraded to a version of portage supporting new eclasses, or patched.
408     The former being much more preferable then the latter for the portage devs.</p>
409     <p>Beyond that, an appropriate window for old eclasses to exist in the tree must be
410 vapier 1.3 determined, and prior to that window passing, an ebuild must be added to the tree
411 g2boojum 1.1 so users can get the old eclasses if needed.</p>
412     <p>For eclass devs to migrate from old to new, it is possible for them to just
413     transfer the old eclass into an appropriate grouping in the new eclass directory,
414 vapier 1.3 although it's advisable they cleanse all cruft out of the eclass. You can
415 g2boojum 1.1 migrate ebuilds gradually over to the new eclass, and don't have to worry about
416     having to support ebuilds from X years back.</p>
417     <p>Essentially, you have a chance to nail the design perfectly/cleanly, and have a
418     window in which to redesign it. It's humbly suggested eclass devs take
419     advantage of it. :)</p>
420     </div>
421     </div>
422     <div class="section" id="backwards-compatibility">
423     <h1><a class="toc-backref" href="#id12" name="backwards-compatibility">Backwards Compatibility</a></h1>
424 g2boojum 1.2 <p>All backwards compatibility issues are addressed in line, but a recap is offered-
425     it's suggested that if the a particular compatibility issue is
426 g2boojum 1.1 questioned/worried over, the reader read the relevant section. There should be
427     a more in depth discussion of the issue, along with a more extensive explanation
428 g2boojum 1.2 of the potential solutions, and reasons for the chosen solution.</p>
429 g2boojum 1.1 <p>To recap:</p>
430     <pre class="literal-block">
431     New eclasses and elib functionality will be tied to a specific portage
432 vapier 1.3 version. A DEPENDs on said portage version should address this for rsync
433 g2boojum 1.1 users who refuse to upgrade to a portage version that supports the new
434     eclasses/elibs and will gradually be unable to merge ebuilds that use said
435     functionality. It is their choice to upgrade, as such, the gradual
436     'thinning' of available ebuilds should they block the portage upgrade is
437     their responsibility.
438    
439     Old eclasses at some point in the future should be removed from the tree,
440 vapier 1.3 and released in a tarball/ebuild. This will cause installed ebuilds that
441 g2boojum 1.2 rely on the old eclass to be unable to unmerge, with the same applying for
442     merging of binpkgs dependent on the following paragraph.
443 g2boojum 1.1
444 g2boojum 1.2 The old eclass-compat is only required for users who do not upgrade their
445     portage installation, and one further exemption- if the user has somehow
446     corrupted/destroyed their installed pkgs database (/var/db/pkg currently),
447     in the process, they've lost their saved environments. The eclass-compat
448     ebuild would be required for ebuilds that required older eclasses in such a
449     case. Note, this case is rare also- as clarified above, it's mentioned
450     strictly to be complete, it's not much of a real world scenario as elaborated
451     above.
452 g2boojum 1.1 </pre>
453     </div>
454     <div class="section" id="copyright">
455     <h1><a class="toc-backref" href="#id13" name="copyright">Copyright</a></h1>
456     <p>This document has been placed in the public domain.</p>
457     </div>
458 vapier 1.3
459 g2boojum 1.1 </div>
460 vapier 1.3 <div class="footer">
461 g2boojum 1.1 <hr class="footer" />
462     <a class="reference" href="glep-0033.txt">View document source</a>.
463 vapier 1.3 Generated on: 2005-09-15 02:37 UTC.
464 g2boojum 1.1 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.
465 vapier 1.3
466 g2boojum 1.1 </div>
467     </body>
468     </html>

  ViewVC Help
Powered by ViewVC 1.1.20