1 |
On Fri, 16 Sep 2016 17:20:28 +0200 |
2 |
Ulrich Mueller <ulm@g.o> wrote: |
3 |
|
4 |
> > 3. Each entry in FILES dictates that the given file is "Used by" the |
5 |
> > ebuild. |
6 |
> |
7 |
> Do you mean "file" in its Unix meaning here, i.e. including |
8 |
> directories? Or only regular files? |
9 |
|
10 |
I'd say regular files for now. Mostly because I think it should be explicit |
11 |
if you want to include directory children. |
12 |
|
13 |
/expressions/ could be expanded during MD5 generation against ${REPO}.../files/ |
14 |
path, which would probably work in the case of "${PV}/*" for instance. |
15 |
|
16 |
But I'm undecided if I like that, only suggesting to get people thinking about |
17 |
it as a possibility. |
18 |
|
19 |
> |
20 |
> > 4. the FILES variable is expanded and interpreted during package |
21 |
> > sourcing |
22 |
> |
23 |
> > 5. "repoman full" and "repoman ci" must ensure any entry listed in FILES |
24 |
> > exists |
25 |
> |
26 |
> > 6. "repoman full" and "repoman ci" must ensure any use of ${FILESDIR} |
27 |
> > without a declared FILES variable is a failure. |
28 |
> |
29 |
> > 7. Under this future EAPI, the location of where ${FILESDIR} expands to |
30 |
> > changes to a location within |
31 |
> > ${PORTAGE_TEMPDIR}/portage/${CATEGORY}/${PF}/files |
32 |
> |
33 |
> > 8. The contents of ${PORTAGE_TEMPDIR}/portage/${CATEGORY}/${PF}/files |
34 |
> > is generated from the source-time $FILES entry by mapping entries from |
35 |
> > ${repostory_location}, either by symlink or by copy (though copy is |
36 |
> > preferred to avoid potential side effects), mapping their heirachy |
37 |
> > exactly ( that is, preserving directory structure ) |
38 |
> |
39 |
> > 9. Files not listed in/indicated by $FILES are thus unavailable to the |
40 |
> > ebuild at runtime and will not be seen in ${FILESDIR}, even if they are |
41 |
> > in ${repository_location} |
42 |
> |
43 |
> AFAICS that proposal goes into a direction which is somewhat opposite |
44 |
> to what we have pursued in EAPI 6. There, we have allowed directories |
45 |
> as arguments to eapply, in order to simplify application of patchsets. |
46 |
> Now maintainers would have to list all single files contained in the |
47 |
> directory again. |
48 |
|
49 |
This wouldn't change the ability for eapply to do that when used with |
50 |
patch-tarballs in SRC_URI, and it wouldn't change the ability to call: |
51 |
|
52 |
eapply ${FILESDIR} |
53 |
|
54 |
It would only change where ${FILESDIR} was and what it contained. |
55 |
|
56 |
But the use of that against directories in $FILEDIR is slightly more |
57 |
cause for concern at present, though mostly due to abuse where files/* |
58 |
are all considered. Such abuses /could/ possibly be made more obvious |
59 |
with this mechanism. |
60 |
|
61 |
> Also I am not sure if I like the additional burden imposed on ebuild |
62 |
> maintainers by requiring double bookkeeping. FILESDIR is already a |
63 |
> well defined location for the support files needed by a package. |
64 |
|
65 |
Agreed to an extent. Though I was of the opinion that to a greater extent |
66 |
at least, such concern cases would simply be relocating the book-keeping |
67 |
logic for defining the file-sets to be at the FILES location, which would |
68 |
give you a canonical variable to consume elsewhere. |
69 |
|
70 |
In theory under the expression system, you could have "FILES=*" which |
71 |
would basically regress the feature to what it currently is, but then having |
72 |
the feature at all wouldn't be very useful. |
73 |
|
74 |
It could be an "opt-in" feature instead of a mandatory one mind, where it would |
75 |
basically degrade to a default mechanic of "Like a PATCHES array, but repoman |
76 |
can check it" |
77 |
|
78 |
> |
79 |
> > :: Benefits :: |
80 |
> |
81 |
> > [...] |
82 |
> |
83 |
> > 4. Due to referential integrity, it should be trivial to identify which |
84 |
> > files are needed by a given ebuild, and which files are now redundant, |
85 |
> > assisting with keeping the tree pruned. |
86 |
> |
87 |
> How does a file in FILESDIR get stale? The typical scenario is that a |
88 |
> patch will no longer be needed after a version bump and pruning of old |
89 |
> ebuild versions. I fear that with FILES the problem would simply be |
90 |
> shifted: instead of forgetting to delete the stale file, people would |
91 |
> forget removing the patch from the FILES list. |
92 |
> |
93 |
> Ulrich |
94 |
|
95 |
I Know those seem like similar problems, but the subtle difference does stick out |
96 |
for me. |
97 |
|
98 |
With the existing scenario the basic assumption is that *every* ebuild |
99 |
needs a given file, and you have to check every ebuild and make sure |
100 |
you don't break one. |
101 |
|
102 |
And this is true *even* if you're a maintainer who tries to keep things tight. |
103 |
|
104 |
A good maintainer can prune the contents of an ebuilds FILES= list *at |
105 |
the time they perform the bump*, where as by contrast, you *cant* prune |
106 |
the files themselves at the time of the bump, because they are still |
107 |
needed. |
108 |
|
109 |
So in a sense, the use of a FILES= list allows you to keep things |
110 |
contextualised, you can mark which you still need, and remove that |
111 |
which you don't. |
112 |
|
113 |
Which means when a second maintainer comes along and removes an ebuild, |
114 |
they don't need to have the knowledge the first maintainer internalised |
115 |
when they made the bump. |
116 |
|
117 |
Removing the ebuild will immediately alert them of the presense of |
118 |
unused files, because they will cease to be indicated by any ebuilds. |
119 |
|
120 |
So you're keeping the knowledge of "I changed the ebuild" attached to |
121 |
"I changed which files are needed", which will be known at the same |
122 |
time. |
123 |
|
124 |
And portage then communicates that knowledge through time to the person |
125 |
who removes the old ebuilds, so they know which files are no longer |
126 |
neeed, without having to check every existing ebuild. |
127 |
|
128 |
Obviously my entire suggestion is not complete and there's probably a |
129 |
few details that need to be sorted out before it approaches |
130 |
properly-useful. |
131 |
|
132 |
And maybe it will serve us better if it starts out as an elective |
133 |
instead of a mandatory feature. |