# git grep '(' = segfault 9 messages from 2009-04-27 to 2009-04-27. Participants: Dmitry Potapov, Linus Torvalds, Michał Kiedrowicz, Erik Faye-Lund, Matthieu Moy, Junio C Hamano. Thread: https://gitlist.dev/t/19089 ## Dmitry Potapov, 2009-04-27 17:46 Subject: git grep '(' = segfault Message-ID: <37fcd2780904271046r7740ed42t3c9438e7aa93374@mail.gmail.com> URL: https://gitlist.dev/e/37fcd2780904271046r7740ed42t3c9438e7aa93374%40mail.gmail.com ``` Hi, I have tried a few version of Git and got "Segmentation fault" when run: $ git grep '(' I am not very familiar with grep code and don't have time to dig into it right now. So, maybe someone else can take a look at it. Thanks, Dmitry ``` ## Linus Torvalds, 2009-04-27 18:10 Subject: Re: git grep '(' = segfault Message-ID: URL: https://gitlist.dev/e/alpine.LFD.2.00.0904271059060.22156%40localhost.localdomain In-Reply-To: <37fcd2780904271046r7740ed42t3c9438e7aa93374@mail.gmail.com> ``` On Mon, 27 Apr 2009, Dmitry Potapov wrote: > > I have tried a few version of Git and got "Segmentation fault" when > run: > > $ git grep '(' > > I am not very familiar with grep code and don't have time to dig > into it right now. So, maybe someone else can take a look at it. Good job. Something like this should fix it. Linus --- grep.c | 8 ++++++-- 1 files changed, 6 insertions(+), 2 deletions(-) diff --git a/grep.c b/grep.c index f3a27d7..04c777a 100644 --- a/grep.c +++ b/grep.c @@ -72,6 +72,8 @@ static struct grep_expr *compile_pattern_atom(struct grep_pat **list) struct grep_expr *x; p = *list; + if (!p) + return NULL; switch (p->token) { case GREP_PATTERN: /* atom */ case GREP_PATTERN_HEAD: @@ -84,8 +86,6 @@ static struct grep_expr *compile_pattern_atom(struct grep_pat **list) case GREP_OPEN_PAREN: *list = p->next; x = compile_pattern_or(list); - if (!x) - return NULL; if (!*list || (*list)->token != GREP_CLOSE_PAREN) die("unmatched parenthesis"); *list = (*list)->next; @@ -101,6 +101,8 @@ static struct grep_expr *compile_pattern_not(struct grep_pat **list) struct grep_expr *x; p = *list; + if (!p) + return NULL; switch (p->token) { case GREP_NOT: if (!p->next) @@ -372,6 +374,8 @@ static int match_expr_eval(struct grep_expr *x, char *bol, char *eol, int h = 0; regmatch_t match; + if (!x) + die("Not a valid grep expression"); switch (x->node) { case GREP_NODE_ATOM: h = match_one_pattern(x->u.atom, bol, eol, ctx, &match, 0); ``` ## Michał Kiedrowicz, 2009-04-27 18:15 Subject: Re: git grep '(' = segfault Message-ID: <20090427201513.5d598e3c@gmail.com> URL: https://gitlist.dev/e/20090427201513.5d598e3c%40gmail.com In-Reply-To: <37fcd2780904271046r7740ed42t3c9438e7aa93374@mail.gmail.com> ``` Dmitry Potapov wrote: > Hi, > > I have tried a few version of Git and got "Segmentation fault" when > run: > > $ git grep '(' > > I am not very familiar with grep code and don't have time to dig > into it right now. So, maybe someone else can take a look at it. > > Thanks, > Dmitry > -- Try this patch, however I'm not sure if it is correct. --- diff --git a/grep.c b/grep.c index f3a27d7..5748355 100644 --- a/grep.c +++ b/grep.c @@ -84,10 +84,10 @@ static struct grep_expr *compile_pattern_atom(struct grep_pat **list) case GREP_OPEN_PAREN: *list = p->next; x = compile_pattern_or(list); - if (!x) - return NULL; if (!*list || (*list)->token != GREP_CLOSE_PAREN) die("unmatched parenthesis"); + if (!x) + return NULL; *list = (*list)->next; return x; default: ``` ## Erik Faye-Lund, 2009-04-27 18:42 Subject: Re: git grep '(' = segfault Message-ID: <40aa078e0904271142u71d7ba97w9df80667c0d9b955@mail.gmail.com> URL: https://gitlist.dev/e/40aa078e0904271142u71d7ba97w9df80667c0d9b955%40mail.gmail.com In-Reply-To: ``` On Mon, Apr 27, 2009 at 8:10 PM, Linus Torvalds wrote: > Something like this should fix it. > >                Linus (Sorry that you're getting this mail twice, Linus - I forgot to reply all) This does indeed fix the crash, but I find the resulting operation quite bizarre. I'm getting a "fatal: unmatched paranthesis", however "git grep '(('" is happy to simply grep for the double parenthesis. Now, if I had done "git grep '\('", I'd expect an unmatched parenthesis-error... Am I misunderstanding something here? $ git grep '(' fatal: unmatched parenthesis $ git grep '((' Documentation/CodingGuidelines: - We use Arithmetic Expansion $(( ... )). [...] $ git grep '\(' fatal: command line, '\(': Unmatched ( or \( -- Erik "kusma" Faye-Lund kusmabite@gmail.com (+47) 986 59 656 ``` ## Linus Torvalds, 2009-04-27 18:50 Subject: Re: git grep '(' = segfault Message-ID: URL: https://gitlist.dev/e/alpine.LFD.2.00.0904271149480.22156%40localhost.localdomain In-Reply-To: <40aa078e0904271142u71d7ba97w9df80667c0d9b955@mail.gmail.com> ``` On Mon, 27 Apr 2009, Erik Faye-Lund wrote: > On Mon, Apr 27, 2009 at 8:10 PM, Linus Torvalds > wrote: > > Something like this should fix it. > > > >                Linus > > (Sorry that you're getting this mail twice, Linus - I forgot to reply all) Payback time. Now _you'll_ get it twice. Mmmwwhahahahaahaaa! > $ git grep '(' > fatal: unmatched parenthesis Try this: git grep -e '(' because 'git grep' actually allows multiple patterns, so you can do things like git grep '(' -e hello --or -e 'hi there' ')' --and -e world Practical? Maybe. Linus ``` ## Erik Faye-Lund, 2009-04-27 18:53 Subject: Re: git grep '(' = segfault Message-ID: <40aa078e0904271153s48fccf1dj4b1bff62abbaf049@mail.gmail.com> URL: https://gitlist.dev/e/40aa078e0904271153s48fccf1dj4b1bff62abbaf049%40mail.gmail.com In-Reply-To: ``` > Practical? Maybe. Confusing? Yes. ;) -- Erik "kusma" Faye-Lund kusmabite@gmail.com (+47) 986 59 656 ``` ## Matthieu Moy, 2009-04-27 18:54 Subject: Re: git grep '(' = segfault Message-ID: URL: https://gitlist.dev/e/vpqfxfund0m.fsf%40bauges.imag.fr In-Reply-To: <40aa078e0904271142u71d7ba97w9df80667c0d9b955@mail.gmail.com> ``` Erik Faye-Lund writes: > $ git grep '(' > fatal: unmatched parenthesis try this: git grep -e '(' The idea is that you can do things like git grep -e foo --and '(' -e bar --or -e boz ')' to build a boolean expression of expressions. -- Matthieu ``` ## Erik Faye-Lund, 2009-04-27 19:02 Subject: Re: git grep '(' = segfault Message-ID: <40aa078e0904271202ye89ebf2wf417c623d8fe5689@mail.gmail.com> URL: https://gitlist.dev/e/40aa078e0904271202ye89ebf2wf417c623d8fe5689%40mail.gmail.com In-Reply-To: ``` Yeah, thanks for explaining. But isn't this functionality kind of redundant since we have the -E switch? As far as I can see, "git grep -E "(bar|boz)" should do the same thing as "git grep -e foo --and '(' -e bar --or -e boz ')'"... On Mon, Apr 27, 2009 at 8:54 PM, Matthieu Moy wrote: > Erik Faye-Lund writes: > >> $ git grep '(' >> fatal: unmatched parenthesis > > try this: git grep -e '(' > > The idea is that you can do things like > > git grep -e foo --and '(' -e bar --or -e boz ')' > > to build a boolean expression of expressions. > > -- > Matthieu > -- Erik "kusma" Faye-Lund kusmabite@gmail.com (+47) 986 59 656 ``` ## Junio C Hamano, 2009-04-27 23:18 Subject: Re: git grep '(' = segfault Message-ID: <7vd4axwurv.fsf@gitster.siamese.dyndns.org> URL: https://gitlist.dev/e/7vd4axwurv.fsf%40gitster.siamese.dyndns.org In-Reply-To: <40aa078e0904271202ye89ebf2wf417c623d8fe5689@mail.gmail.com> ``` Erik Faye-Lund writes: > Yeah, thanks for explaining. But isn't this functionality kind of > redundant since we have the -E switch? > As far as I can see, "git grep -E "(bar|boz)" should do the same thing > as "git grep -e foo --and '(' -e bar --or -e boz ')'"... With -E, --or is not strictly necessary. With only -E, however, it is cumbersome to express what --and can. Once you learn the handiness of --and, naturally you would start wanting to be able to say --or (especially if you do not even know about -E). ```