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

Parent Directory Parent Directory | Revision Log Revision Log

Revision 1.4 - (hide annotations) (download) (as text)
Thu Sep 15 21:04:13 2005 UTC (13 years, 2 months ago) by g2boojum
Branch: MAIN
Changes since 1.3: +34 -29 lines
File MIME type: text/html

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
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 g2boojum 1.4 <tr class="field"><th class="field-name">Version:</th><td class="field-body">1.5</td>
36 g2boojum 1.1 </tr>
37 g2boojum 1.4 <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">2005/09/15 21:02:11</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 g2boojum 1.4 <tr class="field"><th class="field-name">Status:</th><td class="field-body">Approved</td>
42 g2boojum 1.1 </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 g2boojum 1.4 <tr class="field"><th class="field-name">Post-History:</th><td class="field-body">29-Jan-2005 6-Mar-2005 15-Sep-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 g2boojum 1.4 <li><a class="reference" href="#status" id="id2" name="id2">Status</a></li>
58     <li><a class="reference" href="#abstract" id="id3" name="id3">Abstract</a></li>
59     <li><a class="reference" href="#terminology" id="id4" name="id4">Terminology</a></li>
60     <li><a class="reference" href="#motivation-and-rationale" id="id5" name="id5">Motivation and Rationale</a></li>
61     <li><a class="reference" href="#specification" id="id6" name="id6">Specification</a><ul>
62     <li><a class="reference" href="#ebuild-libraries-elibs-for-short" id="id7" name="id7">Ebuild Libraries (elibs for short)</a></li>
63     <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>
64     <li><a class="reference" href="#the-end-of-backwards-compatibility" id="id9" name="id9">The end of backwards compatibility...</a></li>
65     <li><a class="reference" href="#tree-restructuring" id="id10" name="id10">Tree restructuring</a></li>
66     <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>
67     <li><a class="reference" href="#migrating-to-the-new-setup" id="id12" name="id12">Migrating to the new setup</a></li>
68 g2boojum 1.1 </ul>
69     </li>
70 g2boojum 1.4 <li><a class="reference" href="#backwards-compatibility" id="id13" name="id13">Backwards Compatibility</a></li>
71     <li><a class="reference" href="#copyright" id="id14" name="id14">Copyright</a></li>
72 g2boojum 1.1 </ul>
73     </div>
74 g2boojum 1.4 <div class="section" id="status">
75     <h1><a class="toc-backref" href="#id2" name="status">Status</a></h1>
76     <p>Approved by the Gentoo Council on 15 September 2005.</p>
77     </div>
78 g2boojum 1.1 <div class="section" id="abstract">
79 g2boojum 1.4 <h1><a class="toc-backref" href="#id3" name="abstract">Abstract</a></h1>
80 g2boojum 1.1 <p>For any design, the transition from theoretical to applied exposes inadequacies
81     in the original design. This document is intended to document, and propose a
82     revision of the current eclass setup to address current eclass inadequacies.</p>
83 g2boojum 1.2 <p>This document proposes several things- the creation of ebuild libraries, 'elibs',
84 g2boojum 1.1 a narrowing of the focus of eclasses, a move of eclasses w/in the tree, the
85     addition of changelogs, and a way to allow for simple eclass gpg signing.
86     In general, a large scale restructuring of what eclasses are and how they're
87     implemented. Essentially version two of the eclass setup.</p>
88     </div>
89     <div class="section" id="terminology">
90 g2boojum 1.4 <h1><a class="toc-backref" href="#id4" name="terminology">Terminology</a></h1>
91 g2boojum 1.1 <p>From this point on, the proposed eclass setup will be called 'new eclasses', the
92     existing crop (as of this writing) will be referenced as 'old eclasses'. The
93 g2boojum 1.2 distinction is elaborated on within this document.</p>
94 g2boojum 1.1 </div>
95     <div class="section" id="motivation-and-rationale">
96 g2boojum 1.4 <h1><a class="toc-backref" href="#id5" name="motivation-and-rationale">Motivation and Rationale</a></h1>
97 g2boojum 1.1 <p>Eclasses within the tree currently are a bit of a mess- they're forced to
98 g2boojum 1.2 maintain backwards compatibility w/ all previous functionality. In effect,
99 g2boojum 1.1 their api is constant, and can only be added to- never changing the existing
100 vapier 1.3 functionality. This obviously is quite limiting, and leads to cruft accruing in
101 g2boojum 1.1 eclasses as a eclasses design is refined. This needs to be dealt with prior to
102 g2boojum 1.2 eclass code reaching a critical mass where they become unmanageable/fragile
103     (recent pushes for eclass versioning could be interpreted as proof of this).</p>
104 g2boojum 1.1 <p>Beyond that, eclasses were originally intended as a method to allow for ebuilds
105     to use a pre-existing block of code, rather then having to duplicate the code in
106     each ebuild. This is a good thing, but there are ill effects that result from
107 vapier 1.3 the current design. Eclasses inherit other eclasses to get a single function- in
108 g2boojum 1.1 doing so, modifying the the exported 'template' (default src_compile, default
109 vapier 1.3 src_unpack, various vars, etc). All the eclass designer was after was reusing a
110 g2boojum 1.1 function, not making their eclass sensitive to changes in the template of the
111     eclass it's inheriting. The eclass designer -should- be aware of changes in the
112     function they're using, but shouldn't have to worry about their default src_*
113     and pkg_* functions being overwritten, let alone the env changes.</p>
114     <p>Addressing up front why a collection of eclass refinements are being rolled into
115     a single set of changes, parts of this proposal -could- be split into multiple
116     phases. Why do it though? It's simpler for developers to know that the first
117     eclass specification was this, and that the second specification is that,
118     rather then requiring them to be aware of what phase of eclass changes is in
119     progress.</p>
120     <p>By rolling all changes into one large change, a line is intentionally drawn in
121     the sand. Old eclasses allowed for this, behaved this way. New eclasses allow
122 vapier 1.3 for that, and behave this way. This should reduce misconceptions about what is
123 g2boojum 1.1 allowed/possible with eclasses, thus reducing bugs that result from said
124     misconceptions.</p>
125 g2boojum 1.2 <p>A few words on elibs- think of them as a clear definition between behavioral
126     functionality of an eclass, and the library functionality. Eclass's modify
127     template data, and are the basis for other ebuilds- elibs, however are <em>just</em>
128     common bash functionality.</p>
129     <p>Consider the majority of the portage bin/* scripts- these all are candidates for
130     being added to the tree as elibs, as is the bulk of eutils.</p>
131 g2boojum 1.1 </div>
132     <div class="section" id="specification">
133 g2boojum 1.4 <h1><a class="toc-backref" href="#id6" name="specification">Specification</a></h1>
134 g2boojum 1.1 <p>The various parts of this proposal are broken down into a set of changes and
135     elaborations on why a proposed change is preferable. It's advisable to the
136     reader that this be read serially, rather then jumping around.</p>
137     <div class="section" id="ebuild-libraries-elibs-for-short">
138 g2boojum 1.4 <h2><a class="toc-backref" href="#id7" name="ebuild-libraries-elibs-for-short">Ebuild Libraries (elibs for short)</a></h2>
139 g2boojum 1.1 <p>As briefly touched upon in Motivation and Rationale, the original eclass design
140     allowed for the eclass to modify the metadata of an ebuild, metadata being the
141     DEPENDS, RDEPENDS, SRC_URI, IUSE, etc, vars that are required to be constant,
142     and used by portage for dep resolution, fetching, etc. Using the earlier
143     example, if you're after a single function from an eclass (say epatch from
144     eutils), you -don't- want the metadata modifications the eclass you're
145     inheriting might do. You want to treat the eclass you're pulling from as a
146     library, pure and simple.</p>
147     <p>A new directory named elib should be added to the top level of the tree to serve
148     as a repository of ebuild function libraries. Rather then relying on using the
149     source command, an 'elib' function should be added to portage to import that
150     libraries functionality. The reason for the indirection via the function is
151     mostly related to portage internals, but it does serve as an abstraction such
152 g2boojum 1.2 that (for example) zsh compatibility hacks could be hidden in the elib function.</p>
153 g2boojum 1.1 <p>Elib's will be collections of bash functions- they're not allowed to do anything
154     in the global scope aside from function definition, and any -minimal-
155     initialization of the library that is absolutely needed. Additionally, they
156 g2boojum 1.2 cannot modify any ebuild template functions- src_compile, src_unpack. Since they are
157 g2boojum 1.1 required to not modify the metadata keys, nor in any way affect the ebuild aside
158     from providing functionality, they can be conditionally pulled in. They also
159     are allowed to pull in other elibs, but strictly just elibs- no eclasses, just
160 vapier 1.3 other elibs. A real world example would be the eutils eclass.</p>
161 g2boojum 1.1 <p>Portage, since the elib's don't modify metadata, isn't required to track elibs
162     as it tracks eclasses. Thus a change in an elib doesn't result in half the tree
163     forced to be regenerated/marked stale when changed (this is more of an infra
164     benefit, although regen's that take too long due to eclass changes have been
165 g2boojum 1.2 known to cause rsync issues due to missing timestamps).</p>
166     <p>Elibs will not be available in the global scope of an eclass, or ebuild- nor during the
167 vapier 1.3 depends phase (basically a phase that sources the ebuild, to get its metadata). Elib
168 g2boojum 1.2 calls in the global scope will be tracked, but the elib will not be loaded till just before
169 vapier 1.3 the setup phase (pkg_setup). There are two reasons for this- first, it ensures elibs are
170 g2boojum 1.2 completely incapable of modifying metadata. There is no room for confusion, late loading
171     of elibs gives you the functionality for all phases, except for depends- depends being the
172     only phase that is capable of specifying metadata. Second, as an added bonus, late
173     loading reduces the amount of bash sourced for a regen- faster regens. This however is minor,
174     and is an ancillary benefit of the first reason.</p>
175     <p>There are a few further restrictions with elibs--mainly, elibs to load can only be specified
176     in either global scope, or in the setup, unpack, compile, test, and install phases. You can
177     not load elibs in prerm, postrm, preinst, and postinst. The reason being, for *rm phases,
178     installed pkgs will have to look to the tree for the elib, which allows for api drift to cause
179     breakage. For *inst phases, same thing, except the culprit is binpkgs.</p>
180     <p>There is a final restriction--elibs cannot change their exported api dependent on the api
181     (as some eclass do for example). The reason mainly being that elibs are loaded once--not
182     multiple times, as eclasses are.</p>
183     <p>To clarify, for example this is invalid.</p>
184     <pre class="literal-block">
185     if [[ -n ${SOME_VAR} ]]; then
186     func x() { echo &quot;I'm accessible only via tweaking some var&quot;;}
187     else
188     func x() { echo &quot;this is invalid, do not do it.&quot;; }
189     fi
190     </pre>
191 g2boojum 1.1 <p>Regarding maintainability of elibs, it should be a less of a load then old
192     eclasses. One of the major issues with old eclasses is that their functions are
193 vapier 1.3 quite incestuous- they're bound tightly to the env they're defined in. This
194 g2boojum 1.1 makes eclass functions a bit fragile- the restrictions on what can, and cannot
195     be done in elibs will address this, making functionality less fragile (thus a
196     bit more maintainable).</p>
197 g2boojum 1.2 <p>There is no need for backwards compatibility with elibs- they just must work
198 g2boojum 1.1 against the current tree. Thus elibs can be removed when the tree no longer
199     needs them. The reasons for this are explained below.</p>
200     <p>Structuring of the elibs directory will be exactly the same as that of the new
201     eclass directory (detailed below), sans a different extension.</p>
202 g2boojum 1.2 <p>As to why their are so many restrictions, the answer is simple- the definition of
203     what elibs are, what they are capable of, and how to use them is nailed down as much as
204     possible to avoid <em>any</em> ambiguity related to them. The intention is to make it clear,
205     such that no misconceptions occur, resulting in bugs.</p>
206 g2boojum 1.1 </div>
207     <div class="section" id="the-reduced-role-of-eclasses-and-a-clarification-of-existing-eclass-requirements">
208 g2boojum 1.4 <h2><a class="toc-backref" href="#id8" 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>
209 g2boojum 1.1 <p>Since elibs are now intended on holding common bash functionality, the focus of
210 vapier 1.3 eclasses should be in defining an appropriate template for ebuilds. For example,
211 g2boojum 1.1 defining common DEPENDS, RDEPENDS, src_compile functions, src_unpack, etc.
212     Additionally, eclasses should pull in any elibs they need for functionality.</p>
213     <p>Eclass functionality that isn't directly related to the metadata, or src_* and
214     pkg_* funcs should be shifted into elibs to allow for maximal code reuse. This
215     however isn't a hard requirement, merely a strongly worded suggestion.</p>
216     <p>Previously, it was 'strongly' suggested by developers to avoid having any code
217     executed in the global scope that wasn't required. This suggestion is now a
218     requirement. Execute only what must be executed in the global scope. Any code
219     executed in the global scope that is related to configuring/building the package
220     must be placed in pkg_setup. Metadata keys (already a rule, but now stated as
221     an absolute requirement to clarify it) <em>must</em> be constant. The results of
222     metadata keys exported from an ebuild on system A, must be <em>exactly</em> the same as
223     the keys exported on system B.</p>
224     <p>If an eclass (or ebuild for that matter) violates this constant requirement, it
225     leads to portage doing the wrong thing for rsync users- for example, wrong deps
226 g2boojum 1.2 pulled in, leading to compilation failure, or dud deps.</p>
227 g2boojum 1.1 <p>If the existing metadata isn't flexible enough for what is required for a
228     package, the parsing of the metadata is changed to address that. Cases where
229     the constant requirement is violated are known, and a select few are allowed-
230     these are exceptions to the rule that are required due to inadequacies in
231 g2boojum 1.2 portage. Any case where it's determined the constant requirement may need to be
232     violated the dev must make it aware to the majority of devs, along with the portage
233     devs. This should be done prior to committing.</p>
234 g2boojum 1.1 <p>It's quite likely there is a way to allow what you're attempting- if you just go
235 g2boojum 1.2 and do it, the rsync users (our user base) suffer the results of compilation
236 g2boojum 1.1 failures and unneeded deps being pulled in.</p>
237     <p>After that stern reminder, back to new eclasses. Defining INHERITED and ECLASS
238     within the eclass is no longer required. Portage already handles those vars if
239     they aren't defined.</p>
240 g2boojum 1.2 <p>As with elibs, it's no longer required that backwards compatibility be maintained
241     indefinitely- compatibility must be maintained against the current tree, but
242 g2boojum 1.1 just that. As such new eclasses (the true distinction of new vs old is
243     elaborated in the next section) can be removed from the tree once they're no
244     longer in use.</p>
245     </div>
246 g2boojum 1.2 <div class="section" id="the-end-of-backwards-compatibility">
247 g2boojum 1.4 <h2><a class="toc-backref" href="#id9" name="the-end-of-backwards-compatibility">The end of backwards compatibility...</a></h2>
248 vapier 1.3 <p>With current eclasses, once the eclass is in use, its api can no longer be
249 g2boojum 1.1 changed, nor can the eclass ever be removed from the tree. This is why we still
250     have <em>ancient</em> eclasses that are completely unused sitting in the tree, for
251 vapier 1.3 example inherit.eclass. The reason for this, not surprisingly, is a portage
252     deficiency: on unmerging an installed ebuild, portage used the eclass from the
253 g2boojum 1.1 current tree.</p>
254     <p>For a real world example of this, if you merged a glibc 2 years back, whatever
255     eclasses it used must still be compatible, or you may not be able to unmerge the
256     older glibc version during an upgrade to a newer version. So either the glibc
257     maintainer is left with the option of leaving people using ancient versions out
258 g2boojum 1.2 in the rain, or maintaining an ever increasing load of backwards compatibility
259 g2boojum 1.1 cruft in any used eclasses.</p>
260     <p>Binpkgs suffer a similar fate. Merging of a binpkg pulls needed eclasses from
261     the tree, so you may not be able to even merge a binpkg if the eclasses api has
262     changed. If the eclass was removed, you can't even merge the binpkg, period.</p>
263     <p>The next major release of portage will address this- the environment that the
264     ebuild was built in already contains the eclasses functions, as such the env can
265     be re-used rather then relying on the eclass. In other words, binpkgs and
266     installed ebuilds will no longer go and pull needed eclasses from the tree,
267     they'll use the 'saved' version of the eclass they were built/merged with.</p>
268 g2boojum 1.2 <p>So the backwards compatibility requirement for users of the next major portage
269 g2boojum 1.1 version (and beyond) isn't required. All the cruft can be dropped.</p>
270 g2boojum 1.2 <p>The problem is that there will be users using older versions of portage that don't
271     support this functionality- these older installations <em>cannot</em> use the
272     new eclasses, due to the fact that their portage version is incapable of
273     properly relying on the env- in other words, the varying api of the eclass will
274     result in user-visible failures during unmerging.</p>
275     <p>So we're able to do a clean break of all old eclasses, and api cruft, but we need
276     a means to basically disallow access to the new eclasses for all portage versions
277     incapable of properly handling the env requirements.</p>
278     <p>Unfortunately, we cannot just rely on a different grouping/naming convention within
279     the old eclass directory. The new eclasses must be inaccessible, and portage throws
280     a snag into this- the existing inherit function that is used to handle existing
281     eclasses. Basically, whatever it's passed (inherit kernel or inherit
282 g2boojum 1.1 kernel/kernel) it will pull in (kernel.eclass, and kernel/kernel.eclass
283     respectively). So even if the new eclasses were implemented within a
284     subdirectory of the eclass dir in the tree, all current portage versions would
285     still be able to access them.</p>
286     <p>In other words, these new eclasses would in effect, be old eclasses since older
287     portage versions could still access them.</p>
288     </div>
289     <div class="section" id="tree-restructuring">
290 g2boojum 1.4 <h2><a class="toc-backref" href="#id10" name="tree-restructuring">Tree restructuring</a></h2>
291 g2boojum 1.1 <p>There are only two way to block the existing (as of this writing) inherit
292     functionality from accessing the new eclasses- either change the extension of
293 g2boojum 1.2 eclasses to something other then 'eclass', or to have them stored in a separate
294 g2boojum 1.1 subdirectory of the tree then eclass.</p>
295     <p>The latter is preferable, and the proposed solution. Reasons are- the current
296     eclass directory is already overgrown. Structuring of the new eclass dir
297     (clarified below) will allow for easier signing, ChangeLogs, and grouping of
298     eclasses. New eclasses allow for something akin to a clean break and have new
299     capabilities/requirements, thus it's advisable to start with a clean directory,
300     devoid of all cruft from the old eclass implementation.</p>
301     <p>If it's unclear as to why the old inherit function <em>cannot</em> access the new
302     eclasses, please reread the previous section. It's unfortunately a requirement
303     to take advantage of all that the next major portage release will allow.</p>
304 g2boojum 1.2 <p>The proposed directory structure is ${PORTDIR}/include/{eclass,elib}.
305 g2boojum 1.1 Something like ${PORTDIR}/new-eclass, or ${PORTDIR}/eclass-ng could be used
306 vapier 1.3 (although many would cringe at the -ng), but such a name is unwise. Consider the
307 g2boojum 1.1 possibility (likely a fact) that new eclasses someday may be found lacking, and
308 vapier 1.3 refined further (version three as it were). Or perhaps we want to add yet more
309 g2boojum 1.1 functionality with direct relation to sourcing new files, and we would then need
310     to further populate ${PORTDIR}.</p>
311     <p>The new-eclass directory will be (at least) 2 levels deep- for example:</p>
312 vapier 1.3 <dl class="docutils">
313 g2boojum 1.1 <dt>::</dt>
314     <dd>kernel/
315     kernel/linux-info.eclass
316     kernel/linux-mod.eclass
317     kernel/kernel-2.6.eclass
318     kernel/kernel-2.4.eclass
319     kernel/ChangeLog
320     kernel/Manifest</dd>
321     </dl>
322     <p>No eclasses will be allowed in the base directory- grouping of new eclasses will
323     be required to help keep things tidy, and for the following reasons. Grouping
324     of eclasses allows for the addition of ChangeLogs that are specific to that
325     group of eclasses, grouping of files/patches as needed, and allows for
326 g2boojum 1.2 saner/easier signing of eclasses- you can just stick a signed
327 g2boojum 1.1 Manifest file w/in that grouping, thus providing the information portage needs
328     to ensure no files are missing, and that nothing has been tainted.</p>
329     <p>The elib directory will be structured in the same way, for the same reasons.</p>
330     <p>Repoman will have to be extended to work within new eclass and elib groups, and
331 vapier 1.3 to handle signing and committing. This is intentional, and a good thing. This
332 g2boojum 1.2 gives repoman the possibility of doing sanity checks on elibs/new eclasses.</p>
333     <p>Note these checks will not prevent developers from doing dumb things with eclass-
334     these checks would only be capable of doing basic sanity checks, such as syntax checks.
335     There is no way to prevent people from doing dumb things (exempting perhaps repeated
336     applications of a cattle prod)- these are strictly automatic checks, akin to repoman's
337     dependency checks.</p>
338     </div>
339     <div class="section" id="the-start-of-a-different-phase-of-backwards-compatibility">
340 g2boojum 1.4 <h2><a class="toc-backref" href="#id11" name="the-start-of-a-different-phase-of-backwards-compatibility">The start of a different phase of backwards compatibility</a></h2>
341 g2boojum 1.2 <p>As clarified above, new eclasses will exist in a separate directory that will be
342 g2boojum 1.1 intentionally inaccessible to the inherit function. As such, users of older
343     portage versions <em>will</em> have to upgrade to merge any ebuild that uses elibs/new
344 vapier 1.3 eclasses. A depend on the next major portage version would transparently handle
345 g2boojum 1.2 this for rsync users.</p>
346 g2boojum 1.1 <p>There still is the issue of users who haven't upgraded to the required portage
347 vapier 1.3 version. This is a minor concern frankly- portage releases include new
348 g2boojum 1.1 functionality, and bug fixes. If they won't upgrade, it's assumed they have
349     their reasons and are big boys, thus able to handle the complications themselves.</p>
350     <p>The real issue is broken envs, whether in binpkgs, or for installed packages.
351     Two options exist- either the old eclasses are left in the tree indefinitely, or
352     they're left for N months, then shifted out of the tree, and into a tarball that
353     can be merged.</p>
354     <p>Shifting them out of the tree is advisable for several reasons- less cruft in
355     the tree, but more importantly the fact that they are not signed (thus an angle
356     for attack). Note that the proposed method of eclass signing doesn't even try
357     to address them. Frankly, it's not worth the effort supporting two variations
358     of eclass signing, when the old eclass setup isn't designed to allow for easy
359     signing.</p>
360     <p>If this approach is taken, then either the old eclasses would have to be merged
361     to an overlay directory's eclass directory (ugly), or to a safe location that
362     portage's inherit function knows to look for (less ugly).</p>
363     <p>For users who do not upgrade within the window of N months while the old
364     eclasses are in the tree, as stated, it's assumed they know what they are doing.
365     If they specifically block the new portage version, as the ebuilds in the tree
366     migrate to the new eclasses, they will have less and less ebuilds available to
367 g2boojum 1.2 them. If they tried injecting the new portage version (lying to portage,
368     essentially), portage would bail since it cannot find the new eclass.
369     For ebuilds that use the new eclasses, there really isn't any way to sidestep
370     the portage version requirement- same as it has been for other portage features.</p>
371 g2boojum 1.1 <p>What is a bit more annoying is that once the old eclasses are out of the tree,
372 g2boojum 1.2 if a user has not upgraded to a portage version supporting env processing, they
373     will lose the ability to unmerge any installed ebuild that used an old
374     eclass. Same cause, different symptom being they will lose the ability to merge
375     any tbz2 that uses old eclasses also.</p>
376     <p>There is one additional case that is a rarity, but should be noted- if a user
377     has suffered significant corruption of their installed package database (vdb). This is
378     ignoring the question of whether the vdb is even usable at this point, but the possibility
379     exists for the saved envs to be non usable due to either A) missing, or B) corrupted.
380     In such a case, even with the new portage capabilities, they would need
381     the old eclass compat ebuild.</p>
382     <p>Note for this to happen requires either rather... unwise uses of root, or significant
383     fs corruption. Regardless of the cause, it's quite likely for this to even become an
384     issue, the system's vdb is completely unusable. It's a moot issue at that point.
385     If you lose your vdb, or it gets seriously damaged, it's akin to lobotomizing portage-
386 vapier 1.3 it doesn't know what's installed, it doesn't know of its own files, and in general,
387 g2boojum 1.2 a rebuilding of the system is about the only sane course of action. The missing env is
388     truly the least of the users concern in such a case.</p>
389     <p>Continuing with the more likely scenario, users unwilling to upgrade portage will
390     <em>not</em> be left out in the rain. Merging the old eclass compat ebuild will provide
391 vapier 1.3 the missing eclasses, thus providing that lost functionality.</p>
392 g2boojum 1.2 <p>Note the intention isn't to force them to upgrade, hence the ability to restore the
393 vapier 1.3 lost functionality. The intention is to clean up the existing mess, and allow us
394     to move forward. The saying &quot;you've got to break a few eggs to make an omelet&quot;
395 g2boojum 1.1 is akin, exempting the fact we're providing a way to make the eggs whole again
396     (the king's men would've loved such an option).</p>
397     </div>
398     <div class="section" id="migrating-to-the-new-setup">
399 g2boojum 1.4 <h2><a class="toc-backref" href="#id12" name="migrating-to-the-new-setup">Migrating to the new setup</a></h2>
400 g2boojum 1.1 <p>As has been done in the past whenever a change in the tree results in ebuilds
401     requiring a specific version of portage, as ebuilds migrate to the new eclasses,
402     they should depend on a version of portage that supports it. From the users
403     viewpoint, this transparently handles the migration.</p>
404     <p>This isn't so transparent for devs or a particular infrastructure server however.
405     Devs, due to them using cvs for their tree, lack the pregenerated cache rsync
406     users have. Devs will have to be early adopters of the new portage. Older
407     portage versions won't be able to access the new eclasses, thus the local cache
408     generation for that ebuild will fail, ergo the depends on a newer portage
409     version won't transparently handle it for them.</p>
410     <p>Additionally, prior to any ebuilds in the tree using the new eclasses, the
411     infrastructure server that generates the cache for rsync users will have to
412     either be upgraded to a version of portage supporting new eclasses, or patched.
413     The former being much more preferable then the latter for the portage devs.</p>
414     <p>Beyond that, an appropriate window for old eclasses to exist in the tree must be
415 vapier 1.3 determined, and prior to that window passing, an ebuild must be added to the tree
416 g2boojum 1.1 so users can get the old eclasses if needed.</p>
417     <p>For eclass devs to migrate from old to new, it is possible for them to just
418     transfer the old eclass into an appropriate grouping in the new eclass directory,
419 vapier 1.3 although it's advisable they cleanse all cruft out of the eclass. You can
420 g2boojum 1.1 migrate ebuilds gradually over to the new eclass, and don't have to worry about
421     having to support ebuilds from X years back.</p>
422     <p>Essentially, you have a chance to nail the design perfectly/cleanly, and have a
423     window in which to redesign it. It's humbly suggested eclass devs take
424     advantage of it. :)</p>
425     </div>
426     </div>
427     <div class="section" id="backwards-compatibility">
428 g2boojum 1.4 <h1><a class="toc-backref" href="#id13" name="backwards-compatibility">Backwards Compatibility</a></h1>
429 g2boojum 1.2 <p>All backwards compatibility issues are addressed in line, but a recap is offered-
430     it's suggested that if the a particular compatibility issue is
431 g2boojum 1.1 questioned/worried over, the reader read the relevant section. There should be
432     a more in depth discussion of the issue, along with a more extensive explanation
433 g2boojum 1.2 of the potential solutions, and reasons for the chosen solution.</p>
434 g2boojum 1.1 <p>To recap:</p>
435     <pre class="literal-block">
436     New eclasses and elib functionality will be tied to a specific portage
437 vapier 1.3 version. A DEPENDs on said portage version should address this for rsync
438 g2boojum 1.1 users who refuse to upgrade to a portage version that supports the new
439     eclasses/elibs and will gradually be unable to merge ebuilds that use said
440     functionality. It is their choice to upgrade, as such, the gradual
441     'thinning' of available ebuilds should they block the portage upgrade is
442     their responsibility.
444     Old eclasses at some point in the future should be removed from the tree,
445 vapier 1.3 and released in a tarball/ebuild. This will cause installed ebuilds that
446 g2boojum 1.2 rely on the old eclass to be unable to unmerge, with the same applying for
447     merging of binpkgs dependent on the following paragraph.
448 g2boojum 1.1
449 g2boojum 1.2 The old eclass-compat is only required for users who do not upgrade their
450     portage installation, and one further exemption- if the user has somehow
451     corrupted/destroyed their installed pkgs database (/var/db/pkg currently),
452     in the process, they've lost their saved environments. The eclass-compat
453     ebuild would be required for ebuilds that required older eclasses in such a
454     case. Note, this case is rare also- as clarified above, it's mentioned
455     strictly to be complete, it's not much of a real world scenario as elaborated
456     above.
457 g2boojum 1.1 </pre>
458     </div>
459     <div class="section" id="copyright">
460 g2boojum 1.4 <h1><a class="toc-backref" href="#id14" name="copyright">Copyright</a></h1>
461 g2boojum 1.1 <p>This document has been placed in the public domain.</p>
462     </div>
463 vapier 1.3
464 g2boojum 1.1 </div>
465 vapier 1.3 <div class="footer">
466 g2boojum 1.1 <hr class="footer" />
467     <a class="reference" href="glep-0033.txt">View document source</a>.
468 g2boojum 1.4 Generated on: 2005-09-15 21:03 UTC.
469 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.
470 vapier 1.3
471 g2boojum 1.1 </div>
472     </body>
473     </html>

  ViewVC Help
Powered by ViewVC 1.1.20