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

Parent Directory Parent Directory | Revision Log Revision Log

Revision 1.8 - (hide annotations) (download) (as text)
Sun Oct 14 17:00:15 2007 UTC (11 years, 5 months ago) by antarus
Branch: MAIN
Changes since 1.7: +4 -251 lines
File MIME type: text/html
the canary on 53 went well, changing the rest

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 antarus 1.8
5 g2boojum 1.1 <head>
6     <meta http-equiv="Content-Type" content="text/html; charset=utf-8" />
7 g2boojum 1.5 <meta name="generator" content="Docutils 0.4: http://docutils.sourceforge.net/" />
8 g2boojum 1.1 <title>GLEP 33 -- Eclass Restructure/Redesign</title>
9 antarus 1.8 <link rel="stylesheet" href="tools/glep.css" type="text/css" />
10 g2boojum 1.1 </head>
11     <body bgcolor="white">
12     <table class="navigation" cellpadding="0" cellspacing="0"
13     width="100%" border="0">
14     <tr><td class="navicon" width="150" height="35">
15     <a href="http://www.gentoo.org/" title="Gentoo Linux Home Page">
16     <img src="http://www.gentoo.org/images/gentoo-new.gif" alt="[Gentoo]"
17     border="0" width="150" height="35" /></a></td>
18     <td class="textlinks" align="left">
19     [<b><a href="http://www.gentoo.org/">Gentoo Linux Home</a></b>]
20 antarus 1.8 [<b><a href="http://www.gentoo.org/proj/en/glep">GLEP Index</a></b>]
21 g2boojum 1.5 [<b><a href="http://www.gentoo.org/proj/en/glep/glep-0033.txt">GLEP Source</a></b>]
22 g2boojum 1.1 </td></tr></table>
23 vapier 1.3 <table class="rfc2822 docutils field-list" frame="void" rules="none">
24 g2boojum 1.1 <col class="field-name" />
25     <col class="field-body" />
26     <tbody valign="top">
27     <tr class="field"><th class="field-name">GLEP:</th><td class="field-body">33</td>
28     </tr>
29     <tr class="field"><th class="field-name">Title:</th><td class="field-body">Eclass Restructure/Redesign</td>
30     </tr>
31 g2boojum 1.5 <tr class="field"><th class="field-name">Version:</th><td class="field-body">1.6</td>
32 g2boojum 1.1 </tr>
33 g2boojum 1.5 <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>
34 g2boojum 1.1 </tr>
35 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>
36 g2boojum 1.1 </tr>
37 g2boojum 1.5 <tr class="field"><th class="field-name">Status:</th><td class="field-body">Moribund</td>
38 g2boojum 1.1 </tr>
39     <tr class="field"><th class="field-name">Type:</th><td class="field-body">Standards Track</td>
40     </tr>
41 g2boojum 1.5 <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>
42 g2boojum 1.1 </tr>
43     <tr class="field"><th class="field-name">Created:</th><td class="field-body">29-Jan-2005</td>
44     </tr>
45 g2boojum 1.5 <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>
46 g2boojum 1.1 </tr>
47     </tbody>
48     </table>
49     <hr />
50 g2boojum 1.5 <div class="contents topic">
51     <p class="topic-title first"><a id="contents" name="contents">Contents</a></p>
52 g2boojum 1.1 <ul class="simple">
53 g2boojum 1.4 <li><a class="reference" href="#status" id="id2" name="id2">Status</a></li>
54     <li><a class="reference" href="#abstract" id="id3" name="id3">Abstract</a></li>
55     <li><a class="reference" href="#terminology" id="id4" name="id4">Terminology</a></li>
56     <li><a class="reference" href="#motivation-and-rationale" id="id5" name="id5">Motivation and Rationale</a></li>
57     <li><a class="reference" href="#specification" id="id6" name="id6">Specification</a><ul>
58     <li><a class="reference" href="#ebuild-libraries-elibs-for-short" id="id7" name="id7">Ebuild Libraries (elibs for short)</a></li>
59     <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>
60     <li><a class="reference" href="#the-end-of-backwards-compatibility" id="id9" name="id9">The end of backwards compatibility...</a></li>
61     <li><a class="reference" href="#tree-restructuring" id="id10" name="id10">Tree restructuring</a></li>
62     <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>
63     <li><a class="reference" href="#migrating-to-the-new-setup" id="id12" name="id12">Migrating to the new setup</a></li>
64 g2boojum 1.1 </ul>
65     </li>
66 g2boojum 1.4 <li><a class="reference" href="#backwards-compatibility" id="id13" name="id13">Backwards Compatibility</a></li>
67     <li><a class="reference" href="#copyright" id="id14" name="id14">Copyright</a></li>
68 g2boojum 1.1 </ul>
69     </div>
70 g2boojum 1.5 <div class="section">
71     <h1><a class="toc-backref" href="#id2" id="status" name="status">Status</a></h1>
72     <p>Approved by the Gentoo Council on 15 September 2005. As of Sept. 2006
73     this GLEP is on hold, pending future revisions.</p>
74     </div>
75     <div class="section">
76     <h1><a class="toc-backref" href="#id3" id="abstract" name="abstract">Abstract</a></h1>
77     <p>For any design, the transition from theoretical to applied exposes inadequacies
78     in the original design. This document is intended to document, and propose a
79 g2boojum 1.1 revision of the current eclass setup to address current eclass inadequacies.</p>
80 g2boojum 1.5 <p>This document proposes several things- the creation of ebuild libraries, 'elibs',
81     a narrowing of the focus of eclasses, a move of eclasses w/in the tree, the
82 g2boojum 1.1 addition of changelogs, and a way to allow for simple eclass gpg signing.
83     In general, a large scale restructuring of what eclasses are and how they're
84     implemented. Essentially version two of the eclass setup.</p>
85     </div>
86 g2boojum 1.5 <div class="section">
87     <h1><a class="toc-backref" href="#id4" id="terminology" name="terminology">Terminology</a></h1>
88 g2boojum 1.1 <p>From this point on, the proposed eclass setup will be called 'new eclasses', the
89     existing crop (as of this writing) will be referenced as 'old eclasses'. The
90 g2boojum 1.2 distinction is elaborated on within this document.</p>
91 g2boojum 1.1 </div>
92 g2boojum 1.5 <div class="section">
93     <h1><a class="toc-backref" href="#id5" id="motivation-and-rationale" name="motivation-and-rationale">Motivation and Rationale</a></h1>
94 g2boojum 1.1 <p>Eclasses within the tree currently are a bit of a mess- they're forced to
95 g2boojum 1.2 maintain backwards compatibility w/ all previous functionality. In effect,
96 g2boojum 1.1 their api is constant, and can only be added to- never changing the existing
97 vapier 1.3 functionality. This obviously is quite limiting, and leads to cruft accruing in
98 g2boojum 1.1 eclasses as a eclasses design is refined. This needs to be dealt with prior to
99 g2boojum 1.2 eclass code reaching a critical mass where they become unmanageable/fragile
100     (recent pushes for eclass versioning could be interpreted as proof of this).</p>
101 g2boojum 1.1 <p>Beyond that, eclasses were originally intended as a method to allow for ebuilds
102     to use a pre-existing block of code, rather then having to duplicate the code in
103     each ebuild. This is a good thing, but there are ill effects that result from
104 vapier 1.3 the current design. Eclasses inherit other eclasses to get a single function- in
105 g2boojum 1.1 doing so, modifying the the exported 'template' (default src_compile, default
106 vapier 1.3 src_unpack, various vars, etc). All the eclass designer was after was reusing a
107 g2boojum 1.1 function, not making their eclass sensitive to changes in the template of the
108     eclass it's inheriting. The eclass designer -should- be aware of changes in the
109     function they're using, but shouldn't have to worry about their default src_*
110     and pkg_* functions being overwritten, let alone the env changes.</p>
111     <p>Addressing up front why a collection of eclass refinements are being rolled into
112     a single set of changes, parts of this proposal -could- be split into multiple
113     phases. Why do it though? It's simpler for developers to know that the first
114 g2boojum 1.5 eclass specification was this, and that the second specification is that,
115     rather then requiring them to be aware of what phase of eclass changes is in
116 g2boojum 1.1 progress.</p>
117     <p>By rolling all changes into one large change, a line is intentionally drawn in
118     the sand. Old eclasses allowed for this, behaved this way. New eclasses allow
119 vapier 1.3 for that, and behave this way. This should reduce misconceptions about what is
120 g2boojum 1.1 allowed/possible with eclasses, thus reducing bugs that result from said
121     misconceptions.</p>
122 g2boojum 1.5 <p>A few words on elibs- think of them as a clear definition between behavioral
123     functionality of an eclass, and the library functionality. Eclass's modify
124     template data, and are the basis for other ebuilds- elibs, however are <em>just</em>
125 g2boojum 1.2 common bash functionality.</p>
126 g2boojum 1.5 <p>Consider the majority of the portage bin/* scripts- these all are candidates for
127 g2boojum 1.2 being added to the tree as elibs, as is the bulk of eutils.</p>
128 g2boojum 1.1 </div>
129 g2boojum 1.5 <div class="section">
130     <h1><a class="toc-backref" href="#id6" id="specification" name="specification">Specification</a></h1>
131 g2boojum 1.1 <p>The various parts of this proposal are broken down into a set of changes and
132     elaborations on why a proposed change is preferable. It's advisable to the
133     reader that this be read serially, rather then jumping around.</p>
134 g2boojum 1.5 <div class="section">
135     <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>
136 g2boojum 1.1 <p>As briefly touched upon in Motivation and Rationale, the original eclass design
137     allowed for the eclass to modify the metadata of an ebuild, metadata being the
138     DEPENDS, RDEPENDS, SRC_URI, IUSE, etc, vars that are required to be constant,
139     and used by portage for dep resolution, fetching, etc. Using the earlier
140     example, if you're after a single function from an eclass (say epatch from
141     eutils), you -don't- want the metadata modifications the eclass you're
142     inheriting might do. You want to treat the eclass you're pulling from as a
143     library, pure and simple.</p>
144     <p>A new directory named elib should be added to the top level of the tree to serve
145     as a repository of ebuild function libraries. Rather then relying on using the
146     source command, an 'elib' function should be added to portage to import that
147 g2boojum 1.5 libraries functionality. The reason for the indirection via the function is
148     mostly related to portage internals, but it does serve as an abstraction such
149 g2boojum 1.2 that (for example) zsh compatibility hacks could be hidden in the elib function.</p>
150 g2boojum 1.1 <p>Elib's will be collections of bash functions- they're not allowed to do anything
151     in the global scope aside from function definition, and any -minimal-
152 g2boojum 1.5 initialization of the library that is absolutely needed. Additionally, they
153 g2boojum 1.2 cannot modify any ebuild template functions- src_compile, src_unpack. Since they are
154 g2boojum 1.1 required to not modify the metadata keys, nor in any way affect the ebuild aside
155     from providing functionality, they can be conditionally pulled in. They also
156     are allowed to pull in other elibs, but strictly just elibs- no eclasses, just
157 vapier 1.3 other elibs. A real world example would be the eutils eclass.</p>
158 g2boojum 1.1 <p>Portage, since the elib's don't modify metadata, isn't required to track elibs
159     as it tracks eclasses. Thus a change in an elib doesn't result in half the tree
160     forced to be regenerated/marked stale when changed (this is more of an infra
161     benefit, although regen's that take too long due to eclass changes have been
162 g2boojum 1.2 known to cause rsync issues due to missing timestamps).</p>
163 g2boojum 1.5 <p>Elibs will not be available in the global scope of an eclass, or ebuild- nor during the
164     depends phase (basically a phase that sources the ebuild, to get its metadata). Elib
165 g2boojum 1.2 calls in the global scope will be tracked, but the elib will not be loaded till just before
166 g2boojum 1.5 the setup phase (pkg_setup). There are two reasons for this- first, it ensures elibs are
167 g2boojum 1.2 completely incapable of modifying metadata. There is no room for confusion, late loading
168 g2boojum 1.5 of elibs gives you the functionality for all phases, except for depends- depends being the
169     only phase that is capable of specifying metadata. Second, as an added bonus, late
170 g2boojum 1.2 loading reduces the amount of bash sourced for a regen- faster regens. This however is minor,
171     and is an ancillary benefit of the first reason.</p>
172     <p>There are a few further restrictions with elibs--mainly, elibs to load can only be specified
173 g2boojum 1.5 in either global scope, or in the setup, unpack, compile, test, and install phases. You can
174     not load elibs in prerm, postrm, preinst, and postinst. The reason being, for *rm phases,
175     installed pkgs will have to look to the tree for the elib, which allows for api drift to cause
176 g2boojum 1.2 breakage. For *inst phases, same thing, except the culprit is binpkgs.</p>
177 g2boojum 1.5 <p>There is a final restriction--elibs cannot change their exported api dependent on the api
178     (as some eclass do for example). The reason mainly being that elibs are loaded once--not
179 g2boojum 1.2 multiple times, as eclasses are.</p>
180     <p>To clarify, for example this is invalid.</p>
181     <pre class="literal-block">
182     if [[ -n ${SOME_VAR} ]]; then
183     func x() { echo &quot;I'm accessible only via tweaking some var&quot;;}
184     else
185     func x() { echo &quot;this is invalid, do not do it.&quot;; }
186     fi
187     </pre>
188 g2boojum 1.1 <p>Regarding maintainability of elibs, it should be a less of a load then old
189     eclasses. One of the major issues with old eclasses is that their functions are
190 vapier 1.3 quite incestuous- they're bound tightly to the env they're defined in. This
191 g2boojum 1.1 makes eclass functions a bit fragile- the restrictions on what can, and cannot
192     be done in elibs will address this, making functionality less fragile (thus a
193     bit more maintainable).</p>
194 g2boojum 1.2 <p>There is no need for backwards compatibility with elibs- they just must work
195 g2boojum 1.1 against the current tree. Thus elibs can be removed when the tree no longer
196     needs them. The reasons for this are explained below.</p>
197     <p>Structuring of the elibs directory will be exactly the same as that of the new
198     eclass directory (detailed below), sans a different extension.</p>
199 g2boojum 1.5 <p>As to why their are so many restrictions, the answer is simple- the definition of
200     what elibs are, what they are capable of, and how to use them is nailed down as much as
201 g2boojum 1.2 possible to avoid <em>any</em> ambiguity related to them. The intention is to make it clear,
202     such that no misconceptions occur, resulting in bugs.</p>
203 g2boojum 1.1 </div>
204 g2boojum 1.5 <div class="section">
205     <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>
206 g2boojum 1.1 <p>Since elibs are now intended on holding common bash functionality, the focus of
207 vapier 1.3 eclasses should be in defining an appropriate template for ebuilds. For example,
208 g2boojum 1.1 defining common DEPENDS, RDEPENDS, src_compile functions, src_unpack, etc.
209     Additionally, eclasses should pull in any elibs they need for functionality.</p>
210     <p>Eclass functionality that isn't directly related to the metadata, or src_* and
211     pkg_* funcs should be shifted into elibs to allow for maximal code reuse. This
212     however isn't a hard requirement, merely a strongly worded suggestion.</p>
213     <p>Previously, it was 'strongly' suggested by developers to avoid having any code
214     executed in the global scope that wasn't required. This suggestion is now a
215     requirement. Execute only what must be executed in the global scope. Any code
216     executed in the global scope that is related to configuring/building the package
217     must be placed in pkg_setup. Metadata keys (already a rule, but now stated as
218     an absolute requirement to clarify it) <em>must</em> be constant. The results of
219     metadata keys exported from an ebuild on system A, must be <em>exactly</em> the same as
220     the keys exported on system B.</p>
221     <p>If an eclass (or ebuild for that matter) violates this constant requirement, it
222     leads to portage doing the wrong thing for rsync users- for example, wrong deps
223 g2boojum 1.2 pulled in, leading to compilation failure, or dud deps.</p>
224 g2boojum 1.1 <p>If the existing metadata isn't flexible enough for what is required for a
225     package, the parsing of the metadata is changed to address that. Cases where
226     the constant requirement is violated are known, and a select few are allowed-
227     these are exceptions to the rule that are required due to inadequacies in
228 g2boojum 1.5 portage. Any case where it's determined the constant requirement may need to be
229     violated the dev must make it aware to the majority of devs, along with the portage
230 g2boojum 1.2 devs. This should be done prior to committing.</p>
231 g2boojum 1.1 <p>It's quite likely there is a way to allow what you're attempting- if you just go
232 g2boojum 1.2 and do it, the rsync users (our user base) suffer the results of compilation
233 g2boojum 1.1 failures and unneeded deps being pulled in.</p>
234     <p>After that stern reminder, back to new eclasses. Defining INHERITED and ECLASS
235     within the eclass is no longer required. Portage already handles those vars if
236     they aren't defined.</p>
237 g2boojum 1.2 <p>As with elibs, it's no longer required that backwards compatibility be maintained
238     indefinitely- compatibility must be maintained against the current tree, but
239 g2boojum 1.1 just that. As such new eclasses (the true distinction of new vs old is
240     elaborated in the next section) can be removed from the tree once they're no
241     longer in use.</p>
242     </div>
243 g2boojum 1.5 <div class="section">
244     <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>
245 vapier 1.3 <p>With current eclasses, once the eclass is in use, its api can no longer be
246 g2boojum 1.1 changed, nor can the eclass ever be removed from the tree. This is why we still
247     have <em>ancient</em> eclasses that are completely unused sitting in the tree, for
248 g2boojum 1.5 example inherit.eclass. The reason for this, not surprisingly, is a portage
249 vapier 1.3 deficiency: on unmerging an installed ebuild, portage used the eclass from the
250 g2boojum 1.1 current tree.</p>
251     <p>For a real world example of this, if you merged a glibc 2 years back, whatever
252     eclasses it used must still be compatible, or you may not be able to unmerge the
253     older glibc version during an upgrade to a newer version. So either the glibc
254     maintainer is left with the option of leaving people using ancient versions out
255 g2boojum 1.2 in the rain, or maintaining an ever increasing load of backwards compatibility
256 g2boojum 1.1 cruft in any used eclasses.</p>
257     <p>Binpkgs suffer a similar fate. Merging of a binpkg pulls needed eclasses from
258     the tree, so you may not be able to even merge a binpkg if the eclasses api has
259     changed. If the eclass was removed, you can't even merge the binpkg, period.</p>
260     <p>The next major release of portage will address this- the environment that the
261     ebuild was built in already contains the eclasses functions, as such the env can
262     be re-used rather then relying on the eclass. In other words, binpkgs and
263     installed ebuilds will no longer go and pull needed eclasses from the tree,
264     they'll use the 'saved' version of the eclass they were built/merged with.</p>
265 g2boojum 1.2 <p>So the backwards compatibility requirement for users of the next major portage
266 g2boojum 1.1 version (and beyond) isn't required. All the cruft can be dropped.</p>
267 g2boojum 1.5 <p>The problem is that there will be users using older versions of portage that don't
268     support this functionality- these older installations <em>cannot</em> use the
269     new eclasses, due to the fact that their portage version is incapable of
270 g2boojum 1.2 properly relying on the env- in other words, the varying api of the eclass will
271     result in user-visible failures during unmerging.</p>
272 g2boojum 1.5 <p>So we're able to do a clean break of all old eclasses, and api cruft, but we need
273     a means to basically disallow access to the new eclasses for all portage versions
274 g2boojum 1.2 incapable of properly handling the env requirements.</p>
275 g2boojum 1.5 <p>Unfortunately, we cannot just rely on a different grouping/naming convention within
276 g2boojum 1.2 the old eclass directory. The new eclasses must be inaccessible, and portage throws
277     a snag into this- the existing inherit function that is used to handle existing
278     eclasses. Basically, whatever it's passed (inherit kernel or inherit
279 g2boojum 1.1 kernel/kernel) it will pull in (kernel.eclass, and kernel/kernel.eclass
280     respectively). So even if the new eclasses were implemented within a
281     subdirectory of the eclass dir in the tree, all current portage versions would
282     still be able to access them.</p>
283     <p>In other words, these new eclasses would in effect, be old eclasses since older
284     portage versions could still access them.</p>
285     </div>
286 g2boojum 1.5 <div class="section">
287     <h2><a class="toc-backref" href="#id10" id="tree-restructuring" name="tree-restructuring">Tree restructuring</a></h2>
288 g2boojum 1.1 <p>There are only two way to block the existing (as of this writing) inherit
289     functionality from accessing the new eclasses- either change the extension of
290 g2boojum 1.2 eclasses to something other then 'eclass', or to have them stored in a separate
291 g2boojum 1.1 subdirectory of the tree then eclass.</p>
292     <p>The latter is preferable, and the proposed solution. Reasons are- the current
293     eclass directory is already overgrown. Structuring of the new eclass dir
294     (clarified below) will allow for easier signing, ChangeLogs, and grouping of
295     eclasses. New eclasses allow for something akin to a clean break and have new
296 g2boojum 1.5 capabilities/requirements, thus it's advisable to start with a clean directory,
297 g2boojum 1.1 devoid of all cruft from the old eclass implementation.</p>
298     <p>If it's unclear as to why the old inherit function <em>cannot</em> access the new
299     eclasses, please reread the previous section. It's unfortunately a requirement
300     to take advantage of all that the next major portage release will allow.</p>
301 g2boojum 1.2 <p>The proposed directory structure is ${PORTDIR}/include/{eclass,elib}.
302 g2boojum 1.1 Something like ${PORTDIR}/new-eclass, or ${PORTDIR}/eclass-ng could be used
303 vapier 1.3 (although many would cringe at the -ng), but such a name is unwise. Consider the
304 g2boojum 1.1 possibility (likely a fact) that new eclasses someday may be found lacking, and
305 vapier 1.3 refined further (version three as it were). Or perhaps we want to add yet more
306 g2boojum 1.1 functionality with direct relation to sourcing new files, and we would then need
307     to further populate ${PORTDIR}.</p>
308     <p>The new-eclass directory will be (at least) 2 levels deep- for example:</p>
309 vapier 1.3 <dl class="docutils">
310 g2boojum 1.1 <dt>::</dt>
311     <dd>kernel/
312     kernel/linux-info.eclass
313     kernel/linux-mod.eclass
314     kernel/kernel-2.6.eclass
315     kernel/kernel-2.4.eclass
316     kernel/ChangeLog
317     kernel/Manifest</dd>
318     </dl>
319     <p>No eclasses will be allowed in the base directory- grouping of new eclasses will
320     be required to help keep things tidy, and for the following reasons. Grouping
321     of eclasses allows for the addition of ChangeLogs that are specific to that
322     group of eclasses, grouping of files/patches as needed, and allows for
323 g2boojum 1.2 saner/easier signing of eclasses- you can just stick a signed
324 g2boojum 1.1 Manifest file w/in that grouping, thus providing the information portage needs
325     to ensure no files are missing, and that nothing has been tainted.</p>
326     <p>The elib directory will be structured in the same way, for the same reasons.</p>
327     <p>Repoman will have to be extended to work within new eclass and elib groups, and
328 vapier 1.3 to handle signing and committing. This is intentional, and a good thing. This
329 g2boojum 1.2 gives repoman the possibility of doing sanity checks on elibs/new eclasses.</p>
330 g2boojum 1.5 <p>Note these checks will not prevent developers from doing dumb things with eclass-
331 g2boojum 1.2 these checks would only be capable of doing basic sanity checks, such as syntax checks.
332 g2boojum 1.5 There is no way to prevent people from doing dumb things (exempting perhaps repeated
333 g2boojum 1.2 applications of a cattle prod)- these are strictly automatic checks, akin to repoman's
334     dependency checks.</p>
335     </div>
336 g2boojum 1.5 <div class="section">
337     <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>
338 g2boojum 1.2 <p>As clarified above, new eclasses will exist in a separate directory that will be
339 g2boojum 1.1 intentionally inaccessible to the inherit function. As such, users of older
340     portage versions <em>will</em> have to upgrade to merge any ebuild that uses elibs/new
341 g2boojum 1.5 eclasses. A depend on the next major portage version would transparently handle
342 g2boojum 1.2 this for rsync users.</p>
343 g2boojum 1.1 <p>There still is the issue of users who haven't upgraded to the required portage
344 vapier 1.3 version. This is a minor concern frankly- portage releases include new
345 g2boojum 1.1 functionality, and bug fixes. If they won't upgrade, it's assumed they have
346     their reasons and are big boys, thus able to handle the complications themselves.</p>
347     <p>The real issue is broken envs, whether in binpkgs, or for installed packages.
348     Two options exist- either the old eclasses are left in the tree indefinitely, or
349     they're left for N months, then shifted out of the tree, and into a tarball that
350     can be merged.</p>
351     <p>Shifting them out of the tree is advisable for several reasons- less cruft in
352     the tree, but more importantly the fact that they are not signed (thus an angle
353     for attack). Note that the proposed method of eclass signing doesn't even try
354     to address them. Frankly, it's not worth the effort supporting two variations
355     of eclass signing, when the old eclass setup isn't designed to allow for easy
356     signing.</p>
357     <p>If this approach is taken, then either the old eclasses would have to be merged
358     to an overlay directory's eclass directory (ugly), or to a safe location that
359     portage's inherit function knows to look for (less ugly).</p>
360     <p>For users who do not upgrade within the window of N months while the old
361     eclasses are in the tree, as stated, it's assumed they know what they are doing.
362     If they specifically block the new portage version, as the ebuilds in the tree
363     migrate to the new eclasses, they will have less and less ebuilds available to
364 g2boojum 1.2 them. If they tried injecting the new portage version (lying to portage,
365 g2boojum 1.5 essentially), portage would bail since it cannot find the new eclass.
366     For ebuilds that use the new eclasses, there really isn't any way to sidestep
367 g2boojum 1.2 the portage version requirement- same as it has been for other portage features.</p>
368 g2boojum 1.1 <p>What is a bit more annoying is that once the old eclasses are out of the tree,
369 g2boojum 1.5 if a user has not upgraded to a portage version supporting env processing, they
370 g2boojum 1.2 will lose the ability to unmerge any installed ebuild that used an old
371 g2boojum 1.5 eclass. Same cause, different symptom being they will lose the ability to merge
372 g2boojum 1.2 any tbz2 that uses old eclasses also.</p>
373 g2boojum 1.5 <p>There is one additional case that is a rarity, but should be noted- if a user
374     has suffered significant corruption of their installed package database (vdb). This is
375 g2boojum 1.2 ignoring the question of whether the vdb is even usable at this point, but the possibility
376     exists for the saved envs to be non usable due to either A) missing, or B) corrupted.
377     In such a case, even with the new portage capabilities, they would need
378     the old eclass compat ebuild.</p>
379 g2boojum 1.5 <p>Note for this to happen requires either rather... unwise uses of root, or significant
380     fs corruption. Regardless of the cause, it's quite likely for this to even become an
381 g2boojum 1.2 issue, the system's vdb is completely unusable. It's a moot issue at that point.
382 g2boojum 1.5 If you lose your vdb, or it gets seriously damaged, it's akin to lobotomizing portage-
383     it doesn't know what's installed, it doesn't know of its own files, and in general,
384     a rebuilding of the system is about the only sane course of action. The missing env is
385 g2boojum 1.2 truly the least of the users concern in such a case.</p>
386     <p>Continuing with the more likely scenario, users unwilling to upgrade portage will
387 g2boojum 1.5 <em>not</em> be left out in the rain. Merging the old eclass compat ebuild will provide
388 vapier 1.3 the missing eclasses, thus providing that lost functionality.</p>
389 g2boojum 1.2 <p>Note the intention isn't to force them to upgrade, hence the ability to restore the
390 vapier 1.3 lost functionality. The intention is to clean up the existing mess, and allow us
391     to move forward. The saying &quot;you've got to break a few eggs to make an omelet&quot;
392 g2boojum 1.1 is akin, exempting the fact we're providing a way to make the eggs whole again
393     (the king's men would've loved such an option).</p>
394     </div>
395 g2boojum 1.5 <div class="section">
396     <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>
397 g2boojum 1.1 <p>As has been done in the past whenever a change in the tree results in ebuilds
398     requiring a specific version of portage, as ebuilds migrate to the new eclasses,
399     they should depend on a version of portage that supports it. From the users
400     viewpoint, this transparently handles the migration.</p>
401     <p>This isn't so transparent for devs or a particular infrastructure server however.
402     Devs, due to them using cvs for their tree, lack the pregenerated cache rsync
403     users have. Devs will have to be early adopters of the new portage. Older
404     portage versions won't be able to access the new eclasses, thus the local cache
405     generation for that ebuild will fail, ergo the depends on a newer portage
406     version won't transparently handle it for them.</p>
407     <p>Additionally, prior to any ebuilds in the tree using the new eclasses, the
408     infrastructure server that generates the cache for rsync users will have to
409     either be upgraded to a version of portage supporting new eclasses, or patched.
410     The former being much more preferable then the latter for the portage devs.</p>
411     <p>Beyond that, an appropriate window for old eclasses to exist in the tree must be
412 vapier 1.3 determined, and prior to that window passing, an ebuild must be added to the tree
413 g2boojum 1.1 so users can get the old eclasses if needed.</p>
414     <p>For eclass devs to migrate from old to new, it is possible for them to just
415     transfer the old eclass into an appropriate grouping in the new eclass directory,
416 vapier 1.3 although it's advisable they cleanse all cruft out of the eclass. You can
417 g2boojum 1.1 migrate ebuilds gradually over to the new eclass, and don't have to worry about
418     having to support ebuilds from X years back.</p>
419     <p>Essentially, you have a chance to nail the design perfectly/cleanly, and have a
420     window in which to redesign it. It's humbly suggested eclass devs take
421     advantage of it. :)</p>
422     </div>
423     </div>
424 g2boojum 1.5 <div class="section">
425     <h1><a class="toc-backref" href="#id13" id="backwards-compatibility" name="backwards-compatibility">Backwards Compatibility</a></h1>
426 g2boojum 1.2 <p>All backwards compatibility issues are addressed in line, but a recap is offered-
427     it's suggested that if the a particular compatibility issue is
428 g2boojum 1.1 questioned/worried over, the reader read the relevant section. There should be
429     a more in depth discussion of the issue, along with a more extensive explanation
430 g2boojum 1.2 of the potential solutions, and reasons for the chosen solution.</p>
431 g2boojum 1.1 <p>To recap:</p>
432     <pre class="literal-block">
433     New eclasses and elib functionality will be tied to a specific portage
434 vapier 1.3 version. A DEPENDs on said portage version should address this for rsync
435 g2boojum 1.1 users who refuse to upgrade to a portage version that supports the new
436     eclasses/elibs and will gradually be unable to merge ebuilds that use said
437     functionality. It is their choice to upgrade, as such, the gradual
438     'thinning' of available ebuilds should they block the portage upgrade is
439     their responsibility.
441     Old eclasses at some point in the future should be removed from the tree,
442 vapier 1.3 and released in a tarball/ebuild. This will cause installed ebuilds that
443 g2boojum 1.5 rely on the old eclass to be unable to unmerge, with the same applying for
444 g2boojum 1.2 merging of binpkgs dependent on the following paragraph.
445 g2boojum 1.5
446     The old eclass-compat is only required for users who do not upgrade their
447 g2boojum 1.2 portage installation, and one further exemption- if the user has somehow
448 g2boojum 1.5 corrupted/destroyed their installed pkgs database (/var/db/pkg currently),
449 g2boojum 1.2 in the process, they've lost their saved environments. The eclass-compat
450 g2boojum 1.5 ebuild would be required for ebuilds that required older eclasses in such a
451     case. Note, this case is rare also- as clarified above, it's mentioned
452 g2boojum 1.2 strictly to be complete, it's not much of a real world scenario as elaborated
453     above.
454 g2boojum 1.1 </pre>
455     </div>
456 g2boojum 1.5 <div class="section">
457     <h1><a class="toc-backref" href="#id14" id="copyright" name="copyright">Copyright</a></h1>
458 g2boojum 1.1 <p>This document has been placed in the public domain.</p>
459     </div>
460 vapier 1.3
461 g2boojum 1.1 </div>
462 vapier 1.3 <div class="footer">
463 g2boojum 1.1 <hr class="footer" />
464     <a class="reference" href="glep-0033.txt">View document source</a>.
465 antarus 1.8 Generated on: 2007-10-13 13:39 UTC.
466 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.
467 vapier 1.3
468 g2boojum 1.1 </div>
469     </body>
470     </html>

  ViewVC Help
Powered by ViewVC 1.1.20