/[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 1.10 - (hide annotations) (download) (as text)
Wed Apr 7 22:04:22 2010 UTC (4 years, 8 months ago) by robbat2
Branch: MAIN
CVS Tags: HEAD
Changes since 1.9: +50 -52 lines
File MIME type: text/html
Sync to 1.8 of glep-0033.txt.

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 robbat2 1.10 <meta name="generator" content="Docutils 0.6: http://docutils.sourceforge.net/" />
8 g2boojum 1.1 <title>GLEP 33 -- Eclass Restructure/Redesign</title>
9 robbat2 1.10 <link rel="stylesheet" href="tools/glep.css" type="text/css" /></head>
10 g2boojum 1.1 <body bgcolor="white">
11     <table class="navigation" cellpadding="0" cellspacing="0"
12     width="100%" border="0">
13     <tr><td class="navicon" width="150" height="35">
14     <a href="http://www.gentoo.org/" title="Gentoo Linux Home Page">
15     <img src="http://www.gentoo.org/images/gentoo-new.gif" alt="[Gentoo]"
16     border="0" width="150" height="35" /></a></td>
17     <td class="textlinks" align="left">
18     [<b><a href="http://www.gentoo.org/">Gentoo Linux Home</a></b>]
19 antarus 1.8 [<b><a href="http://www.gentoo.org/proj/en/glep">GLEP Index</a></b>]
20 g2boojum 1.5 [<b><a href="http://www.gentoo.org/proj/en/glep/glep-0033.txt">GLEP Source</a></b>]
21 g2boojum 1.1 </td></tr></table>
22 vapier 1.3 <table class="rfc2822 docutils field-list" frame="void" rules="none">
23 g2boojum 1.1 <col class="field-name" />
24     <col class="field-body" />
25     <tbody valign="top">
26     <tr class="field"><th class="field-name">GLEP:</th><td class="field-body">33</td>
27     </tr>
28     <tr class="field"><th class="field-name">Title:</th><td class="field-body">Eclass Restructure/Redesign</td>
29     </tr>
30 robbat2 1.10 <tr class="field"><th class="field-name">Version:</th><td class="field-body">1.8</td>
31 g2boojum 1.1 </tr>
32 robbat2 1.10 <tr class="field"><th class="field-name">Last-Modified:</th><td class="field-body"><a class="reference external" href="http://www.gentoo.org/cgi-bin/viewcvs.cgi/xml/htdocs/proj/en/glep/glep-0033.txt?cvsroot=gentoo">2010/04/07 22:04:02</a></td>
33 g2boojum 1.1 </tr>
34 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>
35 g2boojum 1.1 </tr>
36 robbat2 1.10 <tr class="field"><th class="field-name">Status:</th><td class="field-body">Moribund</td>
37 g2boojum 1.1 </tr>
38     <tr class="field"><th class="field-name">Type:</th><td class="field-body">Standards Track</td>
39     </tr>
40 robbat2 1.10 <tr class="field"><th class="field-name">Content-Type:</th><td class="field-body"><a class="reference external" href="glep-0002.html">text/x-rst</a></td>
41 g2boojum 1.1 </tr>
42     <tr class="field"><th class="field-name">Created:</th><td class="field-body">29-Jan-2005</td>
43     </tr>
44 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>
45 g2boojum 1.1 </tr>
46     </tbody>
47     </table>
48     <hr />
49 robbat2 1.10 <div class="contents topic" id="contents">
50     <p class="topic-title first">Contents</p>
51 g2boojum 1.1 <ul class="simple">
52 robbat2 1.10 <li><a class="reference internal" href="#status" id="id2">Status</a></li>
53     <li><a class="reference internal" href="#abstract" id="id3">Abstract</a></li>
54     <li><a class="reference internal" href="#terminology" id="id4">Terminology</a></li>
55     <li><a class="reference internal" href="#motivation-and-rationale" id="id5">Motivation and Rationale</a></li>
56     <li><a class="reference internal" href="#specification" id="id6">Specification</a><ul>
57     <li><a class="reference internal" href="#ebuild-libraries-elibs-for-short" id="id7">Ebuild Libraries (elibs for short)</a></li>
58     <li><a class="reference internal" href="#the-reduced-role-of-eclasses-and-a-clarification-of-existing-eclass-requirements" id="id8">The reduced role of Eclasses, and a clarification of existing Eclass requirements</a></li>
59     <li><a class="reference internal" href="#the-end-of-backwards-compatibility" id="id9">The end of backwards compatibility...</a></li>
60     <li><a class="reference internal" href="#tree-restructuring" id="id10">Tree restructuring</a></li>
61     <li><a class="reference internal" href="#the-start-of-a-different-phase-of-backwards-compatibility" id="id11">The start of a different phase of backwards compatibility</a></li>
62     <li><a class="reference internal" href="#migrating-to-the-new-setup" id="id12">Migrating to the new setup</a></li>
63 g2boojum 1.1 </ul>
64     </li>
65 robbat2 1.10 <li><a class="reference internal" href="#backwards-compatibility" id="id13">Backwards Compatibility</a></li>
66     <li><a class="reference internal" href="#copyright" id="id14">Copyright</a></li>
67 g2boojum 1.1 </ul>
68     </div>
69 robbat2 1.10 <div class="section" id="status">
70     <h1><a class="toc-backref" href="#id2">Status</a></h1>
71 g2boojum 1.5 <p>Approved by the Gentoo Council on 15 September 2005. As of Sept. 2006
72     this GLEP is on hold, pending future revisions.</p>
73     </div>
74 robbat2 1.10 <div class="section" id="abstract">
75     <h1><a class="toc-backref" href="#id3">Abstract</a></h1>
76 g2boojum 1.5 <p>For any design, the transition from theoretical to applied exposes inadequacies
77     in the original design. This document is intended to document, and propose a
78 g2boojum 1.1 revision of the current eclass setup to address current eclass inadequacies.</p>
79 g2boojum 1.5 <p>This document proposes several things- the creation of ebuild libraries, 'elibs',
80     a narrowing of the focus of eclasses, a move of eclasses w/in the tree, the
81 g2boojum 1.1 addition of changelogs, and a way to allow for simple eclass gpg signing.
82     In general, a large scale restructuring of what eclasses are and how they're
83     implemented. Essentially version two of the eclass setup.</p>
84     </div>
85 robbat2 1.10 <div class="section" id="terminology">
86     <h1><a class="toc-backref" href="#id4">Terminology</a></h1>
87 g2boojum 1.1 <p>From this point on, the proposed eclass setup will be called 'new eclasses', the
88     existing crop (as of this writing) will be referenced as 'old eclasses'. The
89 g2boojum 1.2 distinction is elaborated on within this document.</p>
90 g2boojum 1.1 </div>
91 robbat2 1.10 <div class="section" id="motivation-and-rationale">
92     <h1><a class="toc-backref" href="#id5">Motivation and Rationale</a></h1>
93 g2boojum 1.1 <p>Eclasses within the tree currently are a bit of a mess- they're forced to
94 g2boojum 1.2 maintain backwards compatibility w/ all previous functionality. In effect,
95 g2boojum 1.1 their api is constant, and can only be added to- never changing the existing
96 vapier 1.3 functionality. This obviously is quite limiting, and leads to cruft accruing in
97 g2boojum 1.1 eclasses as a eclasses design is refined. This needs to be dealt with prior to
98 g2boojum 1.2 eclass code reaching a critical mass where they become unmanageable/fragile
99     (recent pushes for eclass versioning could be interpreted as proof of this).</p>
100 g2boojum 1.1 <p>Beyond that, eclasses were originally intended as a method to allow for ebuilds
101     to use a pre-existing block of code, rather then having to duplicate the code in
102     each ebuild. This is a good thing, but there are ill effects that result from
103 vapier 1.3 the current design. Eclasses inherit other eclasses to get a single function- in
104 g2boojum 1.1 doing so, modifying the the exported 'template' (default src_compile, default
105 vapier 1.3 src_unpack, various vars, etc). All the eclass designer was after was reusing a
106 g2boojum 1.1 function, not making their eclass sensitive to changes in the template of the
107     eclass it's inheriting. The eclass designer -should- be aware of changes in the
108     function they're using, but shouldn't have to worry about their default src_*
109     and pkg_* functions being overwritten, let alone the env changes.</p>
110     <p>Addressing up front why a collection of eclass refinements are being rolled into
111     a single set of changes, parts of this proposal -could- be split into multiple
112     phases. Why do it though? It's simpler for developers to know that the first
113 g2boojum 1.5 eclass specification was this, and that the second specification is that,
114     rather then requiring them to be aware of what phase of eclass changes is in
115 g2boojum 1.1 progress.</p>
116     <p>By rolling all changes into one large change, a line is intentionally drawn in
117     the sand. Old eclasses allowed for this, behaved this way. New eclasses allow
118 vapier 1.3 for that, and behave this way. This should reduce misconceptions about what is
119 g2boojum 1.1 allowed/possible with eclasses, thus reducing bugs that result from said
120     misconceptions.</p>
121 g2boojum 1.5 <p>A few words on elibs- think of them as a clear definition between behavioral
122     functionality of an eclass, and the library functionality. Eclass's modify
123     template data, and are the basis for other ebuilds- elibs, however are <em>just</em>
124 g2boojum 1.2 common bash functionality.</p>
125 g2boojum 1.5 <p>Consider the majority of the portage bin/* scripts- these all are candidates for
126 g2boojum 1.2 being added to the tree as elibs, as is the bulk of eutils.</p>
127 g2boojum 1.1 </div>
128 robbat2 1.10 <div class="section" id="specification">
129     <h1><a class="toc-backref" href="#id6">Specification</a></h1>
130 g2boojum 1.1 <p>The various parts of this proposal are broken down into a set of changes and
131     elaborations on why a proposed change is preferable. It's advisable to the
132     reader that this be read serially, rather then jumping around.</p>
133 robbat2 1.10 <div class="section" id="ebuild-libraries-elibs-for-short">
134     <h2><a class="toc-backref" href="#id7">Ebuild Libraries (elibs for short)</a></h2>
135 g2boojum 1.1 <p>As briefly touched upon in Motivation and Rationale, the original eclass design
136     allowed for the eclass to modify the metadata of an ebuild, metadata being the
137     DEPENDS, RDEPENDS, SRC_URI, IUSE, etc, vars that are required to be constant,
138     and used by portage for dep resolution, fetching, etc. Using the earlier
139     example, if you're after a single function from an eclass (say epatch from
140     eutils), you -don't- want the metadata modifications the eclass you're
141     inheriting might do. You want to treat the eclass you're pulling from as a
142     library, pure and simple.</p>
143     <p>A new directory named elib should be added to the top level of the tree to serve
144     as a repository of ebuild function libraries. Rather then relying on using the
145     source command, an 'elib' function should be added to portage to import that
146 g2boojum 1.5 libraries functionality. The reason for the indirection via the function is
147     mostly related to portage internals, but it does serve as an abstraction such
148 g2boojum 1.2 that (for example) zsh compatibility hacks could be hidden in the elib function.</p>
149 g2boojum 1.1 <p>Elib's will be collections of bash functions- they're not allowed to do anything
150     in the global scope aside from function definition, and any -minimal-
151 g2boojum 1.5 initialization of the library that is absolutely needed. Additionally, they
152 g2boojum 1.2 cannot modify any ebuild template functions- src_compile, src_unpack. Since they are
153 g2boojum 1.1 required to not modify the metadata keys, nor in any way affect the ebuild aside
154     from providing functionality, they can be conditionally pulled in. They also
155     are allowed to pull in other elibs, but strictly just elibs- no eclasses, just
156 vapier 1.3 other elibs. A real world example would be the eutils eclass.</p>
157 g2boojum 1.1 <p>Portage, since the elib's don't modify metadata, isn't required to track elibs
158     as it tracks eclasses. Thus a change in an elib doesn't result in half the tree
159     forced to be regenerated/marked stale when changed (this is more of an infra
160     benefit, although regen's that take too long due to eclass changes have been
161 g2boojum 1.2 known to cause rsync issues due to missing timestamps).</p>
162 g2boojum 1.5 <p>Elibs will not be available in the global scope of an eclass, or ebuild- nor during the
163     depends phase (basically a phase that sources the ebuild, to get its metadata). Elib
164 g2boojum 1.2 calls in the global scope will be tracked, but the elib will not be loaded till just before
165 g2boojum 1.5 the setup phase (pkg_setup). There are two reasons for this- first, it ensures elibs are
166 g2boojum 1.2 completely incapable of modifying metadata. There is no room for confusion, late loading
167 g2boojum 1.5 of elibs gives you the functionality for all phases, except for depends- depends being the
168     only phase that is capable of specifying metadata. Second, as an added bonus, late
169 g2boojum 1.2 loading reduces the amount of bash sourced for a regen- faster regens. This however is minor,
170     and is an ancillary benefit of the first reason.</p>
171     <p>There are a few further restrictions with elibs--mainly, elibs to load can only be specified
172 g2boojum 1.5 in either global scope, or in the setup, unpack, compile, test, and install phases. You can
173     not load elibs in prerm, postrm, preinst, and postinst. The reason being, for *rm phases,
174     installed pkgs will have to look to the tree for the elib, which allows for api drift to cause
175 g2boojum 1.2 breakage. For *inst phases, same thing, except the culprit is binpkgs.</p>
176 g2boojum 1.5 <p>There is a final restriction--elibs cannot change their exported api dependent on the api
177     (as some eclass do for example). The reason mainly being that elibs are loaded once--not
178 g2boojum 1.2 multiple times, as eclasses are.</p>
179     <p>To clarify, for example this is invalid.</p>
180     <pre class="literal-block">
181     if [[ -n ${SOME_VAR} ]]; then
182     func x() { echo &quot;I'm accessible only via tweaking some var&quot;;}
183     else
184     func x() { echo &quot;this is invalid, do not do it.&quot;; }
185     fi
186     </pre>
187 g2boojum 1.1 <p>Regarding maintainability of elibs, it should be a less of a load then old
188     eclasses. One of the major issues with old eclasses is that their functions are
189 vapier 1.3 quite incestuous- they're bound tightly to the env they're defined in. This
190 g2boojum 1.1 makes eclass functions a bit fragile- the restrictions on what can, and cannot
191     be done in elibs will address this, making functionality less fragile (thus a
192     bit more maintainable).</p>
193 g2boojum 1.2 <p>There is no need for backwards compatibility with elibs- they just must work
194 g2boojum 1.1 against the current tree. Thus elibs can be removed when the tree no longer
195     needs them. The reasons for this are explained below.</p>
196     <p>Structuring of the elibs directory will be exactly the same as that of the new
197     eclass directory (detailed below), sans a different extension.</p>
198 g2boojum 1.5 <p>As to why their are so many restrictions, the answer is simple- the definition of
199     what elibs are, what they are capable of, and how to use them is nailed down as much as
200 g2boojum 1.2 possible to avoid <em>any</em> ambiguity related to them. The intention is to make it clear,
201     such that no misconceptions occur, resulting in bugs.</p>
202 g2boojum 1.1 </div>
203 robbat2 1.10 <div class="section" id="the-reduced-role-of-eclasses-and-a-clarification-of-existing-eclass-requirements">
204     <h2><a class="toc-backref" href="#id8">The reduced role of Eclasses, and a clarification of existing Eclass requirements</a></h2>
205 g2boojum 1.1 <p>Since elibs are now intended on holding common bash functionality, the focus of
206 vapier 1.3 eclasses should be in defining an appropriate template for ebuilds. For example,
207 g2boojum 1.1 defining common DEPENDS, RDEPENDS, src_compile functions, src_unpack, etc.
208     Additionally, eclasses should pull in any elibs they need for functionality.</p>
209     <p>Eclass functionality that isn't directly related to the metadata, or src_* and
210     pkg_* funcs should be shifted into elibs to allow for maximal code reuse. This
211     however isn't a hard requirement, merely a strongly worded suggestion.</p>
212     <p>Previously, it was 'strongly' suggested by developers to avoid having any code
213     executed in the global scope that wasn't required. This suggestion is now a
214     requirement. Execute only what must be executed in the global scope. Any code
215     executed in the global scope that is related to configuring/building the package
216     must be placed in pkg_setup. Metadata keys (already a rule, but now stated as
217     an absolute requirement to clarify it) <em>must</em> be constant. The results of
218     metadata keys exported from an ebuild on system A, must be <em>exactly</em> the same as
219     the keys exported on system B.</p>
220     <p>If an eclass (or ebuild for that matter) violates this constant requirement, it
221     leads to portage doing the wrong thing for rsync users- for example, wrong deps
222 g2boojum 1.2 pulled in, leading to compilation failure, or dud deps.</p>
223 g2boojum 1.1 <p>If the existing metadata isn't flexible enough for what is required for a
224     package, the parsing of the metadata is changed to address that. Cases where
225     the constant requirement is violated are known, and a select few are allowed-
226     these are exceptions to the rule that are required due to inadequacies in
227 g2boojum 1.5 portage. Any case where it's determined the constant requirement may need to be
228     violated the dev must make it aware to the majority of devs, along with the portage
229 g2boojum 1.2 devs. This should be done prior to committing.</p>
230 g2boojum 1.1 <p>It's quite likely there is a way to allow what you're attempting- if you just go
231 g2boojum 1.2 and do it, the rsync users (our user base) suffer the results of compilation
232 g2boojum 1.1 failures and unneeded deps being pulled in.</p>
233     <p>After that stern reminder, back to new eclasses. Defining INHERITED and ECLASS
234     within the eclass is no longer required. Portage already handles those vars if
235     they aren't defined.</p>
236 g2boojum 1.2 <p>As with elibs, it's no longer required that backwards compatibility be maintained
237     indefinitely- compatibility must be maintained against the current tree, but
238 g2boojum 1.1 just that. As such new eclasses (the true distinction of new vs old is
239     elaborated in the next section) can be removed from the tree once they're no
240     longer in use.</p>
241     </div>
242 robbat2 1.10 <div class="section" id="the-end-of-backwards-compatibility">
243     <h2><a class="toc-backref" href="#id9">The end of backwards compatibility...</a></h2>
244 vapier 1.3 <p>With current eclasses, once the eclass is in use, its api can no longer be
245 g2boojum 1.1 changed, nor can the eclass ever be removed from the tree. This is why we still
246     have <em>ancient</em> eclasses that are completely unused sitting in the tree, for
247 g2boojum 1.5 example inherit.eclass. The reason for this, not surprisingly, is a portage
248 vapier 1.3 deficiency: on unmerging an installed ebuild, portage used the eclass from the
249 g2boojum 1.1 current tree.</p>
250     <p>For a real world example of this, if you merged a glibc 2 years back, whatever
251     eclasses it used must still be compatible, or you may not be able to unmerge the
252     older glibc version during an upgrade to a newer version. So either the glibc
253     maintainer is left with the option of leaving people using ancient versions out
254 g2boojum 1.2 in the rain, or maintaining an ever increasing load of backwards compatibility
255 g2boojum 1.1 cruft in any used eclasses.</p>
256     <p>Binpkgs suffer a similar fate. Merging of a binpkg pulls needed eclasses from
257     the tree, so you may not be able to even merge a binpkg if the eclasses api has
258     changed. If the eclass was removed, you can't even merge the binpkg, period.</p>
259     <p>The next major release of portage will address this- the environment that the
260     ebuild was built in already contains the eclasses functions, as such the env can
261     be re-used rather then relying on the eclass. In other words, binpkgs and
262     installed ebuilds will no longer go and pull needed eclasses from the tree,
263     they'll use the 'saved' version of the eclass they were built/merged with.</p>
264 g2boojum 1.2 <p>So the backwards compatibility requirement for users of the next major portage
265 g2boojum 1.1 version (and beyond) isn't required. All the cruft can be dropped.</p>
266 g2boojum 1.5 <p>The problem is that there will be users using older versions of portage that don't
267     support this functionality- these older installations <em>cannot</em> use the
268     new eclasses, due to the fact that their portage version is incapable of
269 g2boojum 1.2 properly relying on the env- in other words, the varying api of the eclass will
270     result in user-visible failures during unmerging.</p>
271 g2boojum 1.5 <p>So we're able to do a clean break of all old eclasses, and api cruft, but we need
272     a means to basically disallow access to the new eclasses for all portage versions
273 g2boojum 1.2 incapable of properly handling the env requirements.</p>
274 g2boojum 1.5 <p>Unfortunately, we cannot just rely on a different grouping/naming convention within
275 g2boojum 1.2 the old eclass directory. The new eclasses must be inaccessible, and portage throws
276     a snag into this- the existing inherit function that is used to handle existing
277     eclasses. Basically, whatever it's passed (inherit kernel or inherit
278 g2boojum 1.1 kernel/kernel) it will pull in (kernel.eclass, and kernel/kernel.eclass
279     respectively). So even if the new eclasses were implemented within a
280     subdirectory of the eclass dir in the tree, all current portage versions would
281     still be able to access them.</p>
282     <p>In other words, these new eclasses would in effect, be old eclasses since older
283     portage versions could still access them.</p>
284     </div>
285 robbat2 1.10 <div class="section" id="tree-restructuring">
286     <h2><a class="toc-backref" href="#id10">Tree restructuring</a></h2>
287 g2boojum 1.1 <p>There are only two way to block the existing (as of this writing) inherit
288     functionality from accessing the new eclasses- either change the extension of
289 g2boojum 1.2 eclasses to something other then 'eclass', or to have them stored in a separate
290 g2boojum 1.1 subdirectory of the tree then eclass.</p>
291     <p>The latter is preferable, and the proposed solution. Reasons are- the current
292     eclass directory is already overgrown. Structuring of the new eclass dir
293     (clarified below) will allow for easier signing, ChangeLogs, and grouping of
294     eclasses. New eclasses allow for something akin to a clean break and have new
295 g2boojum 1.5 capabilities/requirements, thus it's advisable to start with a clean directory,
296 g2boojum 1.1 devoid of all cruft from the old eclass implementation.</p>
297     <p>If it's unclear as to why the old inherit function <em>cannot</em> access the new
298     eclasses, please reread the previous section. It's unfortunately a requirement
299     to take advantage of all that the next major portage release will allow.</p>
300 g2boojum 1.2 <p>The proposed directory structure is ${PORTDIR}/include/{eclass,elib}.
301 g2boojum 1.1 Something like ${PORTDIR}/new-eclass, or ${PORTDIR}/eclass-ng could be used
302 vapier 1.3 (although many would cringe at the -ng), but such a name is unwise. Consider the
303 g2boojum 1.1 possibility (likely a fact) that new eclasses someday may be found lacking, and
304 vapier 1.3 refined further (version three as it were). Or perhaps we want to add yet more
305 g2boojum 1.1 functionality with direct relation to sourcing new files, and we would then need
306     to further populate ${PORTDIR}.</p>
307     <p>The new-eclass directory will be (at least) 2 levels deep- for example:</p>
308 vapier 1.3 <dl class="docutils">
309 g2boojum 1.1 <dt>::</dt>
310     <dd>kernel/
311     kernel/linux-info.eclass
312     kernel/linux-mod.eclass
313     kernel/kernel-2.6.eclass
314     kernel/kernel-2.4.eclass
315     kernel/ChangeLog
316     kernel/Manifest</dd>
317     </dl>
318     <p>No eclasses will be allowed in the base directory- grouping of new eclasses will
319     be required to help keep things tidy, and for the following reasons. Grouping
320     of eclasses allows for the addition of ChangeLogs that are specific to that
321     group of eclasses, grouping of files/patches as needed, and allows for
322 g2boojum 1.2 saner/easier signing of eclasses- you can just stick a signed
323 g2boojum 1.1 Manifest file w/in that grouping, thus providing the information portage needs
324     to ensure no files are missing, and that nothing has been tainted.</p>
325     <p>The elib directory will be structured in the same way, for the same reasons.</p>
326     <p>Repoman will have to be extended to work within new eclass and elib groups, and
327 vapier 1.3 to handle signing and committing. This is intentional, and a good thing. This
328 g2boojum 1.2 gives repoman the possibility of doing sanity checks on elibs/new eclasses.</p>
329 g2boojum 1.5 <p>Note these checks will not prevent developers from doing dumb things with eclass-
330 g2boojum 1.2 these checks would only be capable of doing basic sanity checks, such as syntax checks.
331 g2boojum 1.5 There is no way to prevent people from doing dumb things (exempting perhaps repeated
332 g2boojum 1.2 applications of a cattle prod)- these are strictly automatic checks, akin to repoman's
333     dependency checks.</p>
334     </div>
335 robbat2 1.10 <div class="section" id="the-start-of-a-different-phase-of-backwards-compatibility">
336     <h2><a class="toc-backref" href="#id11">The start of a different phase of backwards compatibility</a></h2>
337 g2boojum 1.2 <p>As clarified above, new eclasses will exist in a separate directory that will be
338 g2boojum 1.1 intentionally inaccessible to the inherit function. As such, users of older
339     portage versions <em>will</em> have to upgrade to merge any ebuild that uses elibs/new
340 g2boojum 1.5 eclasses. A depend on the next major portage version would transparently handle
341 g2boojum 1.2 this for rsync users.</p>
342 g2boojum 1.1 <p>There still is the issue of users who haven't upgraded to the required portage
343 vapier 1.3 version. This is a minor concern frankly- portage releases include new
344 g2boojum 1.1 functionality, and bug fixes. If they won't upgrade, it's assumed they have
345     their reasons and are big boys, thus able to handle the complications themselves.</p>
346     <p>The real issue is broken envs, whether in binpkgs, or for installed packages.
347     Two options exist- either the old eclasses are left in the tree indefinitely, or
348     they're left for N months, then shifted out of the tree, and into a tarball that
349     can be merged.</p>
350     <p>Shifting them out of the tree is advisable for several reasons- less cruft in
351     the tree, but more importantly the fact that they are not signed (thus an angle
352     for attack). Note that the proposed method of eclass signing doesn't even try
353     to address them. Frankly, it's not worth the effort supporting two variations
354     of eclass signing, when the old eclass setup isn't designed to allow for easy
355     signing.</p>
356     <p>If this approach is taken, then either the old eclasses would have to be merged
357     to an overlay directory's eclass directory (ugly), or to a safe location that
358     portage's inherit function knows to look for (less ugly).</p>
359     <p>For users who do not upgrade within the window of N months while the old
360     eclasses are in the tree, as stated, it's assumed they know what they are doing.
361     If they specifically block the new portage version, as the ebuilds in the tree
362     migrate to the new eclasses, they will have less and less ebuilds available to
363 g2boojum 1.2 them. If they tried injecting the new portage version (lying to portage,
364 g2boojum 1.5 essentially), portage would bail since it cannot find the new eclass.
365     For ebuilds that use the new eclasses, there really isn't any way to sidestep
366 g2boojum 1.2 the portage version requirement- same as it has been for other portage features.</p>
367 g2boojum 1.1 <p>What is a bit more annoying is that once the old eclasses are out of the tree,
368 g2boojum 1.5 if a user has not upgraded to a portage version supporting env processing, they
369 g2boojum 1.2 will lose the ability to unmerge any installed ebuild that used an old
370 g2boojum 1.5 eclass. Same cause, different symptom being they will lose the ability to merge
371 g2boojum 1.2 any tbz2 that uses old eclasses also.</p>
372 g2boojum 1.5 <p>There is one additional case that is a rarity, but should be noted- if a user
373     has suffered significant corruption of their installed package database (vdb). This is
374 g2boojum 1.2 ignoring the question of whether the vdb is even usable at this point, but the possibility
375     exists for the saved envs to be non usable due to either A) missing, or B) corrupted.
376     In such a case, even with the new portage capabilities, they would need
377     the old eclass compat ebuild.</p>
378 g2boojum 1.5 <p>Note for this to happen requires either rather... unwise uses of root, or significant
379     fs corruption. Regardless of the cause, it's quite likely for this to even become an
380 g2boojum 1.2 issue, the system's vdb is completely unusable. It's a moot issue at that point.
381 g2boojum 1.5 If you lose your vdb, or it gets seriously damaged, it's akin to lobotomizing portage-
382     it doesn't know what's installed, it doesn't know of its own files, and in general,
383     a rebuilding of the system is about the only sane course of action. The missing env is
384 g2boojum 1.2 truly the least of the users concern in such a case.</p>
385     <p>Continuing with the more likely scenario, users unwilling to upgrade portage will
386 g2boojum 1.5 <em>not</em> be left out in the rain. Merging the old eclass compat ebuild will provide
387 vapier 1.3 the missing eclasses, thus providing that lost functionality.</p>
388 g2boojum 1.2 <p>Note the intention isn't to force them to upgrade, hence the ability to restore the
389 vapier 1.3 lost functionality. The intention is to clean up the existing mess, and allow us
390     to move forward. The saying &quot;you've got to break a few eggs to make an omelet&quot;
391 g2boojum 1.1 is akin, exempting the fact we're providing a way to make the eggs whole again
392     (the king's men would've loved such an option).</p>
393     </div>
394 robbat2 1.10 <div class="section" id="migrating-to-the-new-setup">
395     <h2><a class="toc-backref" href="#id12">Migrating to the new setup</a></h2>
396 g2boojum 1.1 <p>As has been done in the past whenever a change in the tree results in ebuilds
397     requiring a specific version of portage, as ebuilds migrate to the new eclasses,
398     they should depend on a version of portage that supports it. From the users
399     viewpoint, this transparently handles the migration.</p>
400     <p>This isn't so transparent for devs or a particular infrastructure server however.
401     Devs, due to them using cvs for their tree, lack the pregenerated cache rsync
402     users have. Devs will have to be early adopters of the new portage. Older
403     portage versions won't be able to access the new eclasses, thus the local cache
404     generation for that ebuild will fail, ergo the depends on a newer portage
405     version won't transparently handle it for them.</p>
406     <p>Additionally, prior to any ebuilds in the tree using the new eclasses, the
407     infrastructure server that generates the cache for rsync users will have to
408     either be upgraded to a version of portage supporting new eclasses, or patched.
409     The former being much more preferable then the latter for the portage devs.</p>
410     <p>Beyond that, an appropriate window for old eclasses to exist in the tree must be
411 vapier 1.3 determined, and prior to that window passing, an ebuild must be added to the tree
412 g2boojum 1.1 so users can get the old eclasses if needed.</p>
413     <p>For eclass devs to migrate from old to new, it is possible for them to just
414     transfer the old eclass into an appropriate grouping in the new eclass directory,
415 vapier 1.3 although it's advisable they cleanse all cruft out of the eclass. You can
416 g2boojum 1.1 migrate ebuilds gradually over to the new eclass, and don't have to worry about
417     having to support ebuilds from X years back.</p>
418     <p>Essentially, you have a chance to nail the design perfectly/cleanly, and have a
419     window in which to redesign it. It's humbly suggested eclass devs take
420     advantage of it. :)</p>
421     </div>
422     </div>
423 robbat2 1.10 <div class="section" id="backwards-compatibility">
424     <h1><a class="toc-backref" href="#id13">Backwards Compatibility</a></h1>
425 g2boojum 1.2 <p>All backwards compatibility issues are addressed in line, but a recap is offered-
426     it's suggested that if the a particular compatibility issue is
427 g2boojum 1.1 questioned/worried over, the reader read the relevant section. There should be
428     a more in depth discussion of the issue, along with a more extensive explanation
429 g2boojum 1.2 of the potential solutions, and reasons for the chosen solution.</p>
430 g2boojum 1.1 <p>To recap:</p>
431     <pre class="literal-block">
432     New eclasses and elib functionality will be tied to a specific portage
433 vapier 1.3 version. A DEPENDs on said portage version should address this for rsync
434 g2boojum 1.1 users who refuse to upgrade to a portage version that supports the new
435     eclasses/elibs and will gradually be unable to merge ebuilds that use said
436     functionality. It is their choice to upgrade, as such, the gradual
437     'thinning' of available ebuilds should they block the portage upgrade is
438     their responsibility.
439    
440     Old eclasses at some point in the future should be removed from the tree,
441 vapier 1.3 and released in a tarball/ebuild. This will cause installed ebuilds that
442 g2boojum 1.5 rely on the old eclass to be unable to unmerge, with the same applying for
443 g2boojum 1.2 merging of binpkgs dependent on the following paragraph.
444 g2boojum 1.5
445     The old eclass-compat is only required for users who do not upgrade their
446 g2boojum 1.2 portage installation, and one further exemption- if the user has somehow
447 g2boojum 1.5 corrupted/destroyed their installed pkgs database (/var/db/pkg currently),
448 g2boojum 1.2 in the process, they've lost their saved environments. The eclass-compat
449 g2boojum 1.5 ebuild would be required for ebuilds that required older eclasses in such a
450     case. Note, this case is rare also- as clarified above, it's mentioned
451 g2boojum 1.2 strictly to be complete, it's not much of a real world scenario as elaborated
452     above.
453 g2boojum 1.1 </pre>
454     </div>
455 robbat2 1.10 <div class="section" id="copyright">
456     <h1><a class="toc-backref" href="#id14">Copyright</a></h1>
457 g2boojum 1.1 <p>This document has been placed in the public domain.</p>
458     </div>
459 vapier 1.3
460 g2boojum 1.1 </div>
461 vapier 1.3 <div class="footer">
462 g2boojum 1.1 <hr class="footer" />
463 robbat2 1.10 <a class="reference external" href="glep-0033.txt">View document source</a>.
464     Generated on: 2010-04-07 22:04 UTC.
465     Generated by <a class="reference external" href="http://docutils.sourceforge.net/">Docutils</a> from <a class="reference external" href="http://docutils.sourceforge.net/rst.html">reStructuredText</a> source.
466 vapier 1.3
467 g2boojum 1.1 </div>
468     </body>
469     </html>

  ViewVC Help
Powered by ViewVC 1.1.20