/[gentoo]/xml/htdocs/proj/en/glep/glep-0055.txt
Gentoo

Contents of /xml/htdocs/proj/en/glep/glep-0055.txt

Parent Directory Parent Directory | Revision Log Revision Log


Revision 1.5 - (show annotations) (download)
Sun May 17 20:56:53 2009 UTC (5 years, 6 months ago) by peper
Branch: MAIN
Changes since 1.4: +40 -2 lines
File MIME type: text/plain
Add a more detailed description of the performance hit.

1 GLEP: 55
2 Title: Use EAPI-suffixed ebuilds (.ebuild-EAPI)
3 Version: $Revision: 1.4 $
4 Last-Modified: $Date: 2009/05/17 17:00:30 $
5 Author: Piotr JaroszyƄski <peper@gentoo.org>
6 Status: Draft
7 Type: Standards Track
8 Content-Type: text/x-rst
9 Created: 17-Dec-2007
10 Post-History: 17-Dec-2007, 22-Dec-2007, 17-May-2009
11
12 "A little learning is a dangerous thing; drink deep, or taste not the Pierian
13 spring: there shallow draughts intoxicate the brain, and drinking largely
14 sobers us again."
15
16 -- Alexander Pope, An Essay on Criticism
17
18 Abstract
19 ========
20
21 This GLEP proposes usage of EAPI-suffixed file extensions for ebuilds (for
22 example, foo-1.2.3.ebuild-1).
23
24 Problem
25 =======
26
27 The current way of specifying the EAPI in ebuilds is flawed. In order to get the
28 EAPI the package manager needs to source the ebuild, which itself needs the EAPI
29 in the first place. Otherwise it imposes a serious limitation, namely every ebuild,
30 using any of the future EAPIs, will have to be source'able by old package
31 managers and hence there is no way to do any of the following:
32
33 * Change the behaviour of inherit in any way (for example, to extend or change
34 eclass functionality).
35
36 * Add new global scope functions in any sane way.
37
38 * Extend versioning rules in an EAPI - for example, addition of the scm
39 suffix - GLEP54 [#GLEP54]_ or allowing more sensible version formats like
40 ``1-rc1``, ``1-alpha`` etc. to match upstream more closely.
41
42 * Use newer bash features.
43
44
45 Current behaviour
46 =================
47
48 Following subsections show what happens if you introduce any of the mentioned
49 changes in an ebuild and try to install it with portage 2.1.6.13.
50
51 Incompatible change of inherit (e.g. make it look in the package dir too)
52 -------------------------------------------------------------------------
53
54 ``sys-apps/foo-1.ebuild``::
55
56 EAPI="5"
57 inherit "foo"
58
59 DESCRIPTION=""
60 HOMEPAGE=""
61 SRC_URI=""
62 ...
63
64 Result::
65
66 *
67 * ERROR: sys-apps/foo-1 failed.
68 * Call stack:
69 * ebuild.sh, line 1879: Called _source_ebuild
70 * ebuild.sh, line 1818: Called source '/var/lib/gentoo/repositories/peper/sys-apps/foo/foo-1.ebuild'
71 * foo-1.ebuild, line 6: Called inherit 'foo'
72 * ebuild.sh, line 1218: Called die
73 * The specific snippet of code:
74 * [ ! -e "$location" ] && die "${1}.eclass could not be found by inherit()"
75 * The die message:
76 * foo.eclass could not be found by inherit()
77 *
78 * If you need support, post the topmost build error, and the call stack if relevant.
79 * This ebuild is from an overlay: '/var/lib/gentoo/repositories/peper/'
80 *
81
82 !!! All ebuilds that could satisfy "sys-apps/foo" have been masked.
83 !!! One of the following masked packages is required to complete your request:
84 - sys-apps/foo-1 (masked by: corruption)
85
86 Current portage looks for eclasses only in the ``eclass`` directory of a
87 repository. This results in a fatal error and ebuild being masked by corruption
88 - might be pretty confusing to users.
89
90 New global scope function
91 -------------------------
92
93 ``sys-apps/foo-1.ebuild``::
94
95 EAPI="5"
96 new_global_scope_function "foo"
97
98 DESCRIPTION=""
99 HOMEPAGE=""
100 SRC_URI=""
101 ...
102
103 Result::
104
105 /var/lib/gentoo/repositories/peper/sys-apps/foo/foo-1.ebuild: line 7: new_global_scope_function: command not found
106
107 !!! All ebuilds that could satisfy "sys-apps/foo" have been masked.
108 !!! One of the following masked packages is required to complete your request:
109 - sys-apps/foo-1 (masked by: EAPI 5)
110
111 The current version of portage supports EAPI '2'. You must upgrade to a
112 newer version of portage before EAPI masked packages can be installed.
113
114 Not that bad as user is advised to upgrade portage.
115
116 New version format
117 ------------------
118
119 ``sys-apps/foo-2-rc1.ebuild``::
120
121 Invalid ebuild name: /var/lib/gentoo/repositories/peper/sys-apps/foo/foo-2-rc1.ebuild
122
123 emerge: there are no ebuilds to satisfy "sys-apps/foo"
124
125 Not the best error message, especially if there are lots of them.
126
127 Use newer bash features
128 -----------------------
129
130 ``|&`` is a new type of redirection added in bash-4. It cannot be used even in
131 local scope as bash still parses the whole ebuild.
132
133 ``sys-apps/foo-1.ebuild``::
134
135 EAPI="5"
136
137 foo() {
138 echo "foo" |& cat
139 }
140
141 Result::
142
143 /var/lib/gentoo/repositories/peper/sys-apps/foo/foo-1.ebuild: line 8: syntax error near unexpected token `&'
144 /var/lib/gentoo/repositories/peper/sys-apps/foo/foo-1.ebuild: line 8: ` echo "foo" |& cat'
145 *
146 * ERROR: sys-apps/foo-1 failed.
147 * Call stack:
148 * ebuild.sh, line 1879: Called _source_ebuild
149 * ebuild.sh, line 1818: Called die
150 * The specific snippet of code:
151 * source "${EBUILD}" || die "error sourcing ebuild"
152 * The die message:
153 * error sourcing ebuild
154 *
155 * If you need support, post the topmost build error, and the call stack if relevant.
156 * This ebuild is from an overlay: '/var/lib/gentoo/repositories/peper/'
157 * ... done!
158
159 !!! All ebuilds that could satisfy "sys-apps/foo" have been masked.
160 !!! One of the following masked packages is required to complete your request:
161 - sys-apps/foo-1 (masked by: corruption)
162
163 Again, not the best error.
164
165 Abstract solution
166 =================
167
168 A solution to this problem has to lift those limitations and the only way to do
169 it is to make the EAPI of an ebuild available to the package managers in a way
170 that doesn't require them to source the ebuild. Another important requirement is
171 for the solution to be backward compatible, which has the pleasant side-effect
172 of making the solution applicable in the Gentoo tree right away. Opposed to
173 waiting an arbitrary amount of time, which is never long enough anyway, as the
174 issues listed on the common portage problems page - [#PortageProblems]_ - show.
175
176 Proposed solution
177 =================
178
179 The proposed solution is to use EAPI-suffixed file extensions for ebuilds. This
180 allows package managers to trivially read the EAPI from the ebuild filename. It
181 is also backwards compatible, because currently ebuilds are recognised by the
182 ``.ebuild`` file extension and hence EAPI-suffixed ebuilds are simply ignored by
183 the package managers.
184
185
186 Specification
187 =============
188
189 Ebuild filename extension syntax: ``ebuild[-<EAPI>]``, where ``[]`` denotes an
190 optional part, and ``<EAPI>`` is the EAPI of the ebuild.
191
192 The EAPI used by the ebuild is the EAPI included in the filename if it is set.
193 Otherwise the EAPI set inside the ebuild is used, which defaults to 0 (this is
194 the current behaviour).
195
196 Ebuilds with unsupported EAPIs are masked.
197
198 It should be considered an error to set the EAPI both in the filename and in the
199 ebuild.
200
201 Examples:
202
203 * ``pkg-1.ebuild``, no EAPI set inside the ebuild
204 EAPI defaults to 0.
205
206 * ``pkg-2.ebuild-1``, no EAPI set inside the ebuild
207 EAPI 1 is used.
208
209 * ``pkg-3.ebuild-1``, ``EAPI="1"``
210 EAPI set in both places - error.
211
212 Note that it is still not permitted to have more than one ebuild with equal
213 category, package name, and version. Although it would have the advantage of
214 allowing authors to provide backwards compatible ebuilds, it would introduce
215 problems too. The first is the requirement to have strict EAPI ordering, the
216 second is ensuring that all the ebuilds for a single category/package-version
217 are equivalent, i.e. installing any of them has exactly the same effect on a
218 given system.
219
220 Also note that it is not a new restriction. It is already possible to illegally
221 have multiple versions with different EAPIs as e.g. ``1.0 == 1.00 == 1.00-r0``
222 and hence you could have ``foo-1.0.ebuild`` with EAPI X and ``foo-1.00.ebuild``
223 with EAPI Y.
224
225 Summary of ideas
226 ================
227
228 EAPI-suffixed ebuilds (proposed solution)
229 -----------------------------------------
230
231 Properties:
232 * Can be used right away: yes
233 * Hurts performance: no
234
235 Some say it is clear and simple, others that it is ugly and unintuitive.
236
237 EAPI in the filename with one-time extension change
238 ---------------------------------------------------
239
240 One of the proposed filename formats:
241 ``<PKG>-<VER>.eapi-<EAPI>.eb``
242
243 Properties:
244 * Can be used right away: yes
245 * Hurts performance: no
246
247 This is equivalent to the proposed solution.
248
249 Some say it is better because the extension is static.
250
251 Easily fetchable EAPI inside the ebuild
252 ---------------------------------------
253
254 Properties:
255 * Can be used right away: no
256 * Hurts performance: yes
257
258 Cannot be used right away as it would trigger the errors shown in Current
259 behaviour section for old package managers.
260
261 Performance decrease comes from the fact that with version format changes in the
262 picture package managers need EAPI to parse the ebuild's version. That means that merely
263 picking the best version of a package requires loading EAPI (from cache or the
264 ebuild) for each available ebuild.
265
266 Here is more or less how the package manager figures out the best available
267 version for a package with N versions available.
268
269 * EAPI in the filename
270
271 * Read the directory containing the package - readdir()
272 * For each ebuild, read its EAPI and using that parse its version - no I/O
273 * Sort the versions - no I/O
274 * Going down from the highest to the lowest version
275
276 * Get the metadata from cache - 2 x stat() + read()
277 * break if the version is visible
278
279 * EAPI in the ebuild
280
281 * Read the directory containing the package - readdir()
282 * For each ebuild load its metadata from cache to get its EAPI - N x (2 x stat() + read())
283 * Sort the versions - no I/O
284 * Going down from the highest to the lowest version
285
286 * (metadata is already loaded) - no I/O
287 * break if the version is visible - no I/O
288
289 The difference is in for how many versions the package manager needs to hit
290 cache. With EAPI in the ebuild it needs to do that for all versions, with EAPI
291 in the filename it depends on versions visibility.
292 For example, package foo has versions 1, 2, 3, 4, 5 and 6. 6 is masked, 5 is
293 ~arch and 1,2,3 and 4 are arch. Say, the user accepts only arch for this
294 package. With EAPI in the filename it will read metadata only for versions
295 6, 5 and 4. With EAPI in the ebuild it needs to load metadata for all versions.
296
297 It's hard to say what's the avarage case, but surely the worst case scenario
298 (when only the lowest version is visible) is uncommon.
299
300 Easily fetchable EAPI inside the ebuild and one-time extension change
301 ---------------------------------------------------------------------
302
303 Properties:
304 * Can be used right away: yes
305 * Hurts performance: yes
306
307 Performance decrease as described in the previous section.
308
309 Some say it is clear and simple, others that it is confusing and unintuitive,
310 because of the arbitrary format restrictions in what is a bash script otherwise.
311
312 Use different subdirectories for different EAPIs, i.e. cat/pkg/eapiX/
313 ---------------------------------------------------------------------
314
315 Properties:
316 * Can be used right away: yes
317 * Hurts performance: yes
318
319 Performance decrease comes from the fact that it adds several more directory
320 reads.
321
322 Some say that it makes it much harder for maintainers to see what they have.
323
324 References
325 ==========
326
327 .. [#GLEP54] GLEP 54, scm package version suffix
328 (http://glep.gentoo.org/glep-0054.html)
329
330 .. [#PortageProblems] Common portage problems
331 (http://www.gentoo.org/proj/en/portage/doc/common-problems.xml)
332
333 Copyright
334 =========
335
336 This document has been placed in the public domain.
337
338 .. vim: set tw=80 fileencoding=utf-8 spell spelllang=en et :

  ViewVC Help
Powered by ViewVC 1.1.20