1 |
> On 8 Nov 2022, at 01:10, Matt Turner <mattst88@g.o> wrote: |
2 |
> |
3 |
> Noticed on ChromeOS when installing a large number of font packages in |
4 |
> parallel: |
5 |
> |
6 |
> /usr/share/fonts/noto/NotoSerifThai-Regular.ttf#new' from 0004 (------r--) to 2440 (r--r-S---) |
7 |
> * ERROR: media-fonts/ipaex-004.01-r1::chromiumos failed (postinst phase): |
8 |
> * failed to fix font files perms |
9 |
> |
10 |
> The "#new" filename is the hint. Portage uses "#new" suffixes when |
11 |
> copying files to the system, and then renames them to their final |
12 |
> filenames. |
13 |
> |
14 |
> This code was executing while another font was in the process of being |
15 |
> copied to the system. Font packages should just ensure that they install |
16 |
> files with correct permissions to begin with, and all except |
17 |
> media-fonts/x11fonts-jmk already use 0644 permissions. |
18 |
> media-fonts/x11fonts-jmk used 0444 (which was probably fine) until the |
19 |
> previous commit which changes its installed files to 0644. |
20 |
> |
21 |
> Bug: https://bugs.gentoo.org/187774 |
22 |
> Signed-off-by: Matt Turner <mattst88@g.o> |
23 |
> --- |
24 |
> eclass/font.eclass | 6 ------ |
25 |
> 1 file changed, 6 deletions(-) |
26 |
> |
27 |
> diff --git a/eclass/font.eclass b/eclass/font.eclass |
28 |
> index 4970c959f7c..0196755ce3e 100644 |
29 |
> --- a/eclass/font.eclass |
30 |
> +++ b/eclass/font.eclass |
31 |
> @@ -186,12 +186,6 @@ font_src_install() { |
32 |
> # @DESCRIPTION: |
33 |
> # Updates fontcache if !prefix and media-libs/fontconfig installed |
34 |
> _update_fontcache() { |
35 |
> - if [[ -d "${EROOT}"/usr/share/fonts ]] ; then |
36 |
> - # unreadable font files = fontconfig segfaults |
37 |
> - find "${EROOT}"/usr/share/fonts/ -type f '!' -perm 0644 \ |
38 |
> - -exec chmod -v 0644 2>/dev/null {} + || die "failed to fix font files perms" |
39 |
> - fi |
40 |
> - |
41 |
> if [[ -z ${ROOT} ]] ; then |
42 |
> if has_version media-libs/fontconfig ; then |
43 |
> ebegin "Updating global fontcache" |
44 |
> -- |
45 |
|
46 |
Can we put an fperms call in src_install just in case (like the eclass originally had |
47 |
before moved to pkg_postinst)? |