Summary: | fix for incorrect FSF snail mail addresses | ||
---|---|---|---|
Product: | glibc | Reporter: | Paul Eggert <eggert> |
Component: | libc | Assignee: | Not yet assigned to anyone <unassigned> |
Status: | RESOLVED FIXED | ||
Severity: | normal | Flags: | fweimer:
security-
|
Priority: | P2 | ||
Version: | unspecified | ||
Target Milestone: | --- | ||
Host: | Target: | ||
Build: | Last reconfirmed: | ||
Attachments: |
patch for incorrect FSF snail mail addresses
patch for nptl/ChangeLog and localedata/ChangeLog patch for Yacc input and output file Patch for gperf input and output files patch for locale files patch for misc. files, mostly .S files revised patch for gperf input and output files glibc-ports patch for snail mail address patch gmp-mparam.h too |
Description
Paul Eggert
2012-02-07 23:52:10 UTC
Comment on attachment 6201 [details]
patch for incorrect FSF snail mail addresses
changing MIME type from text/plain to application/octet-stream
Some of the changed files came from GNU MP - have the copies there been updated? Some in intl/ are, I think, maintained primarily in gettext - the same question applies. nptl/ and localedata/ have their own ChangeLogs. I'll send a similar bug report to the GNU MP and gettext projects at some point. I'll attach a patch to the two ChangeLogs that Joseph mentioned. Created attachment 6202 [details]
patch for nptl/ChangeLog and localedata/ChangeLog
I have reviewed the patch, as far as one can review a 100000-line patch, and committed most of it (all the changes would individually be considered obvious). I omitted the following parts: * config.sub, config.guess: these files should only ever be updated verbatim from upstream config.git, not changed locally. * intl/plural.[cy], locale/programs/{charmap,locfile}-kw.{h,gperf}: it appears that the generated files in these pairs had just had the same substitution applied as the source files, which is not correct as a I think a regeneration would also change #line directives to reflect that the license notices are now one line shorter. Several files had changes removing blank lines at the end of the file. I left those changes in, but please try to avoid including such unrelated changes in future. You may wish to do a followup to change the files with generated outputs correctly (actually regenerating the output) and to change files that are new since your first patch. Also, perhaps you could share your automation for producing this patch so it can be used for ports, or do a corresponding patch for ports (where one file has the even older 675 Mass Ave address)? Please also update gettext, GMP and libidn upstream with changes corresponding to those made in libc to files from those projects. I've reverted the changes to locale files because they caused errors of the form: ./locales/de_DE:9: unterminated string ./locales/de_DE:9: LC_IDENTIFICATION: unknown character in field `address' Charmap: "ISO-8859-1" Inputfile: "de_DE" Outputdir: "de_DE.ISO-8859-1" failed when running the testsuite. More evidence that even if a patch seems obvious you should test it before submitting it.... (It looks like this comes from "escape_char /" in the locale files ... feel free to resubmit a fixed and tested patch.) WAITING pending a new patch for the cases I mentioned as excluded or reverted (and a patch, or sharing the automation to produce one, for the ports repository). Created attachment 6277 [details]
patch for Yacc input and output file
This is an updated patch for intl/plural.y. It also updates intl/plural.c, the generated file.
Created attachment 6278 [details]
Patch for gperf input and output files
This is a patch for gperf input files, which also patches the corresponding output files.
Created attachment 6279 [details]
patch for locale files
This is a patch for the locale files. It corrects the previous attempt to fix the problem, by doubling each instance of "/" in the address strings.
Created attachment 6280 [details]
patch for misc. files, mostly .S files
This is a patch for other files that I found the Boston address in. As far as I know these files are not imported from elsewhere so they can be fixed here (but then, I don't know exactly which files are imported).
I've added four patches just now. The first three try to address Joseph's review in comments 5 and 6. The fourth address all the files that I didn't otherwise see addressed. I don't use any automation for these patches, other than throwaway Emacs macros, so I'm afraid there's no technology to share. I suppose I could do a patch for ports, but I don't know how to get the ports source files, or how to test the result -- can you please advise? http://sourceware.org/glibc/wiki/GlibcGit describes how to check out ports: git clone git://sourceware.org/git/glibc-ports.git Each ports architecture, plus sysdeps/unix/sysv/linux/generic, has its own ChangeLog, but for this sort of mechanical patch it's probably not worth splitting up the patch itself by architecture. I don't think there are any files in ports that are generated files or imported verbatim from elsewhere. On Fri, 9 Mar 2012, eggert at gnu dot org wrote:
> Patch for gperf input and output files
>
> This is a patch for gperf input files, which also patches the corresponding
> output files.
This appears to duplicate the word "License" in the license notices.
Created attachment 6281 [details]
revised patch for gperf input and output files
oops oops.... I have attached a corrected version of the gperf-related patch, one that doesn't say "License License".
Created attachment 6282 [details]
glibc-ports patch for snail mail address
Thanks for the info about glibc-ports. I'm attaching a patch for glibc-ports.
It's 18,000 lines long, so I'm compressing it.
One way to review it which is a bit less mind-numbing, is to filter it
through this shell command:
grep '^[-+]' | grep -vE '^(--- |\+\+\+ )' | sort | uniq -c | sort -rn
This outputs each unique instance of an old (-) and new (-) line, together
with a count of the number of instances. You can ignore this trailing bit:
1 ---
1 --
as that's an artifact of "git format-patch".
I have reviewed and committed the ports patch. Port maintainers may wish to test their ports build to make sure there are no nonobvious causes of breakage missed in the review, but it seemed best to get this issue completely fixed (well ... it appears sysdeps/mips/mips64/gmp-mparam.h was missed by the patch) rather than leaving it indefinitely to get maintainers of eight different ports, with varying activity levels, to fix their ports individually. I have reviewed, tested and committed the libc patches; the only bit left is, I think, sysdeps/mips/mips64/gmp-mparam.h in ports. Created attachment 6283 [details] patch gmp-mparam.h too > sysdeps/mips/mips64/gmp-mparam.h was missed by the patch Sorry. I used "Boston" to find the copyright notices, and that file is the only one that used "Cambridge". I'm attaching a small proposed patch to clean up that glitch. Thanks, I've committed that last patch. |