* zsh history gets destroyed when running out of disk space @ 2007-06-11 20:22 Michael Prokop 2007-06-21 8:04 ` Frank Terbeck 0 siblings, 1 reply; 5+ messages in thread From: Michael Prokop @ 2007-06-11 20:22 UTC (permalink / raw) To: zsh-workers [-- Attachment #1.1: Type: text/plain, Size: 777 bytes --] Hi, Problem: Zsh truncates the zsh history file if you are running out of disk space (AKA ENOSPC). If you don't have any space left in your $HOME and exit zsh you'll find an empty $HISTFILE left. Reason: The savehistfile() function in Src/hist.c does not handle write errors very smart. Fix: See attached patch (against version 4.3.4). The fix might need some more work though. For example there should be much more and better error handling; Zsh shouldn't exit with a return value of 0 when such serious errors happen; the user should be informed about the reason why it failed. regards, -mika- -- ,'"`. http://www.michael-prokop.at/ ( grml.org -» Linux Live-CD for texttool-users and sysadmins `._,' http://www.grml.org/ [-- Attachment #1.2: zsh_fix_history_enospc.patch --] [-- Type: text/x-diff, Size: 864 bytes --] --- Src/hist.c.orig 2007-06-11 21:02:42.000000000 +0200 +++ Src/hist.c 2007-06-11 21:58:37.000000000 +0200 @@ -2158,6 +2158,11 @@ Histent he; zlong xcurhist = curhist - !!(histactive & HA_ACTIVE); int extended_history = isset(EXTENDEDHISTORY); + int write_file = 1; // by default assume we can write the file + + // make sure we don't write anything in case of an error + if (errno == ENOSPC) + write_file = 0; if (!interact || savehistsiz <= 0 || !hist_ring || (!fn && !(fn = getsparam("HISTFILE")))) @@ -2222,7 +2227,7 @@ #endif } } - if (out) { + if (out && write_file) { for (; he && he->histnum <= xcurhist; he = down_histent(he)) { if ((writeflags & HFILE_SKIPDUPS && he->node.flags & HIST_DUP) || (writeflags & HFILE_SKIPFOREIGN && he->node.flags & HIST_FOREIGN) [-- Attachment #2: Digital signature --] [-- Type: application/pgp-signature, Size: 189 bytes --] ^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: zsh history gets destroyed when running out of disk space 2007-06-11 20:22 zsh history gets destroyed when running out of disk space Michael Prokop @ 2007-06-21 8:04 ` Frank Terbeck 2007-06-22 20:28 ` Peter Stephenson 0 siblings, 1 reply; 5+ messages in thread From: Frank Terbeck @ 2007-06-21 8:04 UTC (permalink / raw) To: zsh-workers Michael Prokop <news@michael-prokop.at>: > Problem: > > Zsh truncates the zsh history file if you are running out of disk > space (AKA ENOSPC). If you don't have any space left in your $HOME > and exit zsh you'll find an empty $HISTFILE left. [...] Hey, I experienced just the same a few days back. So, I also think this should be addressed. Maybe a new option could be introduced, that works similar to the 'check_jobs' option that warns if jobs would be killed. Similarly a 'hist_check_space' option could (if set) warn, if a history file can not be written to disk. In any case, I agree with Michael, that a history file should not be truncated to zero-size if the needed space is not available. In that case it's better, IMHO, to keep the old history file. Regards, Frank -- In protocol design, perfection has been reached not when there is nothing left to add, but when there is nothing left to take away. -- RFC 1925 ^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: zsh history gets destroyed when running out of disk space 2007-06-21 8:04 ` Frank Terbeck @ 2007-06-22 20:28 ` Peter Stephenson 2007-06-22 22:40 ` Michael Prokop 0 siblings, 1 reply; 5+ messages in thread From: Peter Stephenson @ 2007-06-22 20:28 UTC (permalink / raw) To: zsh-workers On Thu, 21 Jun 2007 10:04:54 +0200 Frank Terbeck <ft@bewatermyfriend.org> wrote: > Michael Prokop <news@michael-prokop.at>: > > Problem: > > > > Zsh truncates the zsh history file if you are running out of disk > > space (AKA ENOSPC). If you don't have any space left in your $HOME > > and exit zsh you'll find an empty $HISTFILE left. > > Hey, I experienced just the same a few days back. So, I also think > this should be addressed. The following detects errors during writing better. If you're writing a new file it'll leave it wherever it gets to; if you're replacing an old file it will leave the old file alone. That's about the best we can do. I haven't tested this on a failure, but it should be better than what we've got. Index: Src/hist.c =================================================================== RCS file: /cvsroot/zsh/zsh/Src/hist.c,v retrieving revision 1.66 diff -u -r1.66 hist.c --- Src/hist.c 1 Nov 2006 12:25:22 -0000 1.66 +++ Src/hist.c 22 Jun 2007 20:19:50 -0000 @@ -2158,6 +2158,7 @@ Histent he; zlong xcurhist = curhist - !!(histactive & HA_ACTIVE); int extended_history = isset(EXTENDEDHISTORY); + int ret; if (!interact || savehistsiz <= 0 || !hist_ring || (!fn && !(fn = getsparam("HISTFILE")))) @@ -2242,59 +2243,69 @@ } t = start = he->node.nam; if (extended_history) { - fprintf(out, ": %ld:%ld;", (long)he->stim, - he->ftim? (long)(he->ftim - he->stim) : 0L); + ret = fprintf(out, ": %ld:%ld;", (long)he->stim, + he->ftim? (long)(he->ftim - he->stim) : 0L); } else if (*t == ':') - fputc('\\', out); + ret = fputc('\\', out); - for (; *t; t++) { + for (; ret >= 0 && *t; t++) { if (*t == '\n') - fputc('\\', out); - fputc(*t, out); + if ((ret = fputc('\\', out)) < 0) + break; + if ((ret = fputc(*t, out)) < 0) + break; } - fputc('\n', out); + if (ret < 0 || (ret = fputc('\n', out)) < 0) + break; } - if (start && writeflags & HFILE_USE_OPTIONS) { + if (ret >= 0 && start && writeflags & HFILE_USE_OPTIONS) { struct stat sb; - fflush(out); - if (fstat(fileno(out), &sb) == 0) { - lasthist.fsiz = sb.st_size; - lasthist.mtim = sb.st_mtime; + if ((ret = fflush(out)) >= 0) { + if (fstat(fileno(out), &sb) == 0) { + lasthist.fsiz = sb.st_size; + lasthist.mtim = sb.st_mtime; + } + zsfree(lasthist.text); + lasthist.text = ztrdup(start); } - zsfree(lasthist.text); - lasthist.text = ztrdup(start); - } - fclose(out); - if (tmpfile) { - if (rename(tmpfile, unmeta(fn)) < 0) - zerr("can't rename %s.new to $HISTFILE", fn); - free(tmpfile); } + if (fclose(out) < 0 && ret >= 0) + ret = -1; + if (ret >= 0) { + if (tmpfile) { + if (rename(tmpfile, unmeta(fn)) < 0) + zerr("can't rename %s.new to $HISTFILE", fn); + free(tmpfile); + } - if (writeflags & HFILE_SKIPOLD - && !(writeflags & (HFILE_FAST | HFILE_NO_REWRITE))) { - int remember_histactive = histactive; - - /* Zeroing histactive avoids unnecessary munging of curline. */ - histactive = 0; - /* The NULL leaves HISTFILE alone, preserving fn's value. */ - pushhiststack(NULL, savehistsiz, savehistsiz, -1); - - hist_ignore_all_dups |= isset(HISTSAVENODUPS); - readhistfile(fn, err, 0); - hist_ignore_all_dups = isset(HISTIGNOREALLDUPS); - if (histlinect) - savehistfile(fn, err, 0); + if (writeflags & HFILE_SKIPOLD + && !(writeflags & (HFILE_FAST | HFILE_NO_REWRITE))) { + int remember_histactive = histactive; + + /* Zeroing histactive avoids unnecessary munging of curline. */ + histactive = 0; + /* The NULL leaves HISTFILE alone, preserving fn's value. */ + pushhiststack(NULL, savehistsiz, savehistsiz, -1); + + hist_ignore_all_dups |= isset(HISTSAVENODUPS); + readhistfile(fn, err, 0); + hist_ignore_all_dups = isset(HISTIGNOREALLDUPS); + if (histlinect) + savehistfile(fn, err, 0); - pophiststack(); - histactive = remember_histactive; + pophiststack(); + histactive = remember_histactive; + } } - } else if (err) { + } else + ret = -1; + + if (ret < 0 && err) { if (tmpfile) { - zerr("can't write history file %s.new", fn); + zerr("can't write history file %s.new: %e", fn); free(tmpfile); } else - zerr("can't write history file %s", fn); + zerr("can't write history file %s: %e", fn); } unlockhistfile(fn); -- Peter Stephenson <p.w.stephenson@ntlworld.com> Web page now at http://homepage.ntlworld.com/p.w.stephenson/ ^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: zsh history gets destroyed when running out of disk space 2007-06-22 20:28 ` Peter Stephenson @ 2007-06-22 22:40 ` Michael Prokop 2007-06-23 18:09 ` Peter Stephenson 0 siblings, 1 reply; 5+ messages in thread From: Michael Prokop @ 2007-06-22 22:40 UTC (permalink / raw) To: zsh-workers [-- Attachment #1: Type: text/plain, Size: 1458 bytes --] * Peter Stephenson <p.w.stephenson@ntlworld.com> [20070622 22:31]: > On Thu, 21 Jun 2007 10:04:54 +0200 Frank Terbeck <ft@bewatermyfriend.org> wrote: > > Michael Prokop <news@michael-prokop.at>: > > > Problem: > > > Zsh truncates the zsh history file if you are running out of disk > > > space (AKA ENOSPC). If you don't have any space left in your $HOME > > > and exit zsh you'll find an empty $HISTFILE left. > > Hey, I experienced just the same a few days back. So, I also think > > this should be addressed. > The following detects errors during writing better. If you're writing a > new file it'll leave it wherever it gets to; if you're replacing an old > file it will leave the old file alone. That's about the best we can do. > I haven't tested this on a failure, but it should be better than what > we've got. Verified, works. Exiting shell quits with: | zsh: failed to write history file /home/grml/.zsh_history.new: unknown error 1182546552 and an empty /home/grml/.zsh_history.new file is left. My original /home/grml/.zsh_history is left untouched (no truncating anymore). During a shell session the history file is getting longer only as long as there a free bits left (as you intented to). The error code might be improved though. ;) regards, -mika- -- ,'"`. http://www.michael-prokop.at/ ( grml.org -» Linux Live-CD for texttool-users and sysadmins `._,' http://www.grml.org/ [-- Attachment #2: Digital signature --] [-- Type: application/pgp-signature, Size: 189 bytes --] ^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: zsh history gets destroyed when running out of disk space 2007-06-22 22:40 ` Michael Prokop @ 2007-06-23 18:09 ` Peter Stephenson 0 siblings, 0 replies; 5+ messages in thread From: Peter Stephenson @ 2007-06-23 18:09 UTC (permalink / raw) To: zsh-workers On Sat, 23 Jun 2007 00:40:09 +0200 Michael Prokop <news@michael-prokop.at> wrote: > | zsh: failed to write history file /home/grml/.zsh_history.new: unknown error 1182546552 > > The error code might be improved though. ;) Thanks for trying it. I forgot that %e doesn't use errno directly, it expects an argument (that's because sometimes we save the number and do something else before reporting the error). I've also reset errno to 0 beforehand, on the off chance that one of the possible failures doesn't set errno. I don't think that should happen, but I don't want a stray ENOENT from a previous error confusing the matter. Index: Src/hist.c =================================================================== RCS file: /cvsroot/zsh/zsh/Src/hist.c,v retrieving revision 1.67 diff -u -r1.67 hist.c --- Src/hist.c 22 Jun 2007 20:32:20 -0000 1.67 +++ Src/hist.c 23 Jun 2007 18:05:01 -0000 @@ -2190,6 +2190,7 @@ if (isset(SHAREHISTORY)) extended_history = 1; } + errno = 0; if (writeflags & HFILE_APPEND) { tmpfile = NULL; out = fdopen(open(unmeta(fn), @@ -2302,10 +2303,10 @@ if (ret < 0 && err) { if (tmpfile) { - zerr("failed to write history file %s.new: %e", fn); + zerr("failed to write history file %s.new: %e", fn, errno); free(tmpfile); } else - zerr("failed to write history file %s: %e", fn); + zerr("failed to write history file %s: %e", fn, errno); } unlockhistfile(fn); -- Peter Stephenson <p.w.stephenson@ntlworld.com> Web page now at http://homepage.ntlworld.com/p.w.stephenson/ ^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2007-06-23 18:10 UTC | newest] Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed) -- links below jump to the message on this page -- 2007-06-11 20:22 zsh history gets destroyed when running out of disk space Michael Prokop 2007-06-21 8:04 ` Frank Terbeck 2007-06-22 20:28 ` Peter Stephenson 2007-06-22 22:40 ` Michael Prokop 2007-06-23 18:09 ` Peter Stephenson
Code repositories for project(s) associated with this public inbox https://git.vuxu.org/mirror/zsh/ This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox; as well as URLs for NNTP newsgroup(s).