| Message ID | 18f69b9fb1e0277f26c40609c6fc16e25b9d689a.1546898693.git.yann.morin.1998@free.fr |
|---|---|
| State | Changes Requested |
| Headers | show |
| Series | [01/19] infra/pkg-generic: display MESSAGE before running PRE_HOOKS | expand |
El lun., 7 ene. 2019 a las 23:07, Yann E. MORIN (<yann.morin.1998@free.fr>) escribió: > > Currently, we check that no two packages write to the same files, as a > sanity check. We do so by checking which files were touched since the > end of the build (aka beginning of the installation). > > However, when the packages do install the exact same file, i,e, the > same content, we in fact do not really care what package had provided > said file. > > In the past, we avoided that situation because we were md5sum-inf every > files before and after installation. Anything that changed was new or md5sum-ing 'all files' or 'every file'
diff --git a/support/scripts/check-uniq-files b/support/scripts/check-uniq-files index cf9ea292bc..f42edeb534 100755 --- a/support/scripts/check-uniq-files +++ b/support/scripts/check-uniq-files @@ -36,13 +36,18 @@ def main(): return False file_to_pkg = defaultdict(set) + file_md5 = defaultdict(set) for record in parse_pkg_file_list(args.packages_file_list[0]): file_to_pkg[record['file']].add(record['pkg']) + file_md5[record['file']].add(record['md5']) for file in file_to_pkg: if len(file_to_pkg[file]) == 1: continue + if len(file_md5[file]) == 1: + continue + sys.stderr.write(warn.format(args.type, str_decode(file), ", ".join([str_decode(p) for p in file_to_pkg[file]])))
Currently, we check that no two packages write to the same files, as a sanity check. We do so by checking which files were touched since the end of the build (aka beginning of the installation). However, when the packages do install the exact same file, i,e, the same content, we in fact do not really care what package had provided said file. In the past, we avoided that situation because we were md5sum-inf every files before and after installation. Anything that changed was new or modified, and everything that did not change was not modified (but could have been reinstalled). However, since 7fb6e78254 (core/instrumentation: shave minutes off the build time), we're now using mtimes, and we're in the situation that the exact same file installed by two-or-more packages is reported. In such a situation, it is not very interesting to know what package installed the file, because whatever the ordering, or whatever the subset of said packages, we'd have ended up with the same file anyway. One prominent case where this happens, is the fftw family of packages, that all install different libraries, but install the same set of headers and some common utilities, and they are all identical across the family. As such, when the packages all installed the same content (same md5), do not report the file. Only report it if at least two packages installed a different content. Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Cc: John Keeping <john@metanate.com> Cc: Nicolas Cavallari <nicolas.cavallari@green-communications.fr> Cc: Thomas Petazzoni <thomas.petazzoni@bootlin.com> --- support/scripts/check-uniq-files | 5 +++++ 1 file changed, 5 insertions(+)