uninitialized globals

uninitialized globals

[Karl Dahlke]

Perhaps someone can shed some light on this. 
I received this from a friend of mine, who does some packaging for 
debian.

------------------------------ 
I have been running a self-compiled version of the latest edbrowse 
3.8.9 on 
Debian, compiled with GCC 12.2. It segfaults on startup, and the 
backtrace 
suggests it is in main.c:1926, where it nzFree's sslCerts. This is 
likely 
caused by some hardning options of the Debian packaging infrastructure. 
It cause sslCerts to be not NULL-initialised. I couldn't track down 
exactly 
which option it is, but the simple fix is to NULL out the static pointer 
sslCerts. I could hand in a patch, if you like (or a pull request). At 
the 
moment, this makes Edbrowse unusable on Debian. 
------------------------------

I am rather stunned by this. 
I have been using C since 1980. 
Ever since the original K&R, global uninitialized variables are 0. 
I don't understand how sslCerts could be not zero. 
The "simple fix" I'm guessing is to set it to 0, 
but that's not simple because there are dozens of global variables, and 
some static variables too, that are not initialized, that I count on 
being zero. 
If we got past sslCerts we would just run  into another one. 
We'd have to scan through the entire body of code, some 50 thousand 
lines, to find them all and set them all to 0, 
or, 
we could try to understand how and why sslCerts is not null, 
because I would have said that was impossible. 
I just did an internet search, and site after site after site confirms 
that all such uninitialized variables are 0.

Karl Dahlke

Re: uninitialized globals

[Sebastian Humenda]

[-- Attachment #1: Type: text/plain, Size: 12304 bytes --]

Hi

Pardon, I should have sent it to the list.

Karl Dahlke schrieb am 05.04.2024,  4:29 -0400:
>the latest edbrowse 3.8.9 on Debian, compiled with GCC 12.2. It segfaults on
>startup, and the backtrace suggests it is in main.c:1926, where it nzFree's
[…]
[…]

An excerpt of the build output is below this e-mail; it includes the build
flags.

>I am rather stunned by this. I have been using C since 1980. Ever since the
>original K&R, global uninitialized variables are 0. I don't understand how
[…]

I am bored by C standards and alike. I have been writing MISRA C for several
years and I do remember the checkre requiring us to initialise explicitly
every static (i.e. global) as its initialisation was either depending on the
type or implementation defined, I do not recall.

Here is what nm says:

% nm src/edbrowse | grep 'sslCerts'
0000000000247088 D sslCerts

% nm src/edbrowse | fgrep allowJS
% nm src/edbrowse | fgrep allowJS

0000000000247079 D allowJS

>but that's not simple because there are dozens of global variables, and some
>static variables too, that are not initialized, that I count on being zero.
>If we got past sslCerts we would just run  into another one. We'd have to
[…]

Agreed. I actually just looked whether I can call up my.test.website and with
this set, it worked :). I actually was trying out the QuicJS support.

Cheers
Sebastian

=== Build Output
cc -g -O2 -ffile-prefix-map=/tmp/edbrowse=. -fstack-protector-strong -Wformat -Werror=format-security -Wall -Wno-unused -D_FILE_OFFSET_BITS=64 -DEDBROWSE_ON_LINUX -I/usr/include/x86_64-linux-gnu  -g -ggdb -Wextra -Wdate-time -D_FORTIFY_SOURCE=2  -c -o dbodbc.o dbodbc.c
isup.c: In function ‘decodePostData’:
isup.c:1276:30: warning: ‘c’ may be used uninitialized [-Wmaybe-uninitialized]
 1276 |                         *w++ = c;
      |                         ~~~~~^~~
isup.c:1230:14: note: ‘c’ was declared here
 1230 |         char c;
      |              ^
css.c: In function ‘hashSortCrunch’:
css.c:3222:37: warning: ‘j’ may be used uninitialized [-Wmaybe-uninitialized]
 3222 |                         mark->body[j++] = v->t;
      |                                    ~^~
css.c:3202:16: note: ‘j’ was declared here
 3202 |         int i, j, distinct = 0;
      |                ^
css.c: In function ‘do_rules’:
css.c:3106:9: warning: ‘tn’ may be used uninitialized [-Wmaybe-uninitialized]
 3106 |         set_property_bool_t(tn, "inj$css", true);
      |         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
css.c:2957:20: note: ‘tn’ was declared here
 2957 |         const Tag *tn; // the text node that holds before or after text
      |                    ^~
dbops.c: In function ‘stringDate’:
dbops.c:505:17: warning: ‘__builtin_strncpy’ output may be truncated copying 4 bytes from a string of length 7 [-Wstringop-truncation]
  505 |                 strncpy(buf, buf + 4, 4);
      |                 ^
dbodbc.c: In function ‘retsFromOdbc’:
dbodbc.c:937:30: warning: ‘input_length’ may be used uninitialized [-Wmaybe-uninitialized]
  937 |                         rc = SQLGetData(hstmt, (ushort) (i + 1),
      |                              ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  938 |                                         c_type, q1, input_length,
      |                                         ~~~~~~~~~~~~~~~~~~~~~~~~~
  939 |                                         &output_length);
      |                                         ~~~~~~~~~~~~~~~
dbodbc.c:827:14: note: ‘input_length’ was declared here
  827 |         long input_length, output_length;
      |              ^~~~~~~~~~~~
dbodbc.c:937:30: warning: ‘c_type’ may be used uninitialized [-Wmaybe-uninitialized]
  937 |                         rc = SQLGetData(hstmt, (ushort) (i + 1),
      |                              ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  938 |                                         c_type, q1, input_length,
      |                                         ~~~~~~~~~~~~~~~~~~~~~~~~~
  939 |                                         &output_length);
      |                                         ~~~~~~~~~~~~~~~
dbodbc.c:826:15: note: ‘c_type’ was declared here
  826 |         short c_type;           /* C data type */
      |               ^~~~~~
dbops.c: In function ‘stringTime’:
dbops.c:636:28: warning: ‘c’ may be used uninitialized [-Wmaybe-uninitialized]
  636 |                         if (c == 'A')
      |                            ^
dbops.c:593:14: note: ‘c’ was declared here
  593 |         char c;
      |              ^
css.c: In function ‘cssParseLeft’:
css.c:1336:52: warning: ‘sel2’ may be used uninitialized [-Wmaybe-uninitialized]
 1336 |                                         sel2->next = sel, sel2 = sel;
      |                                         ~~~~~~~~~~~^~~~~
css.c:1258:21: note: ‘sel2’ was declared here
 1258 |         struct sel *sel2;
      |                     ^~~~
dbodbc.c: In function ‘oneRetValue.constprop’:
dbodbc.c:1114:19: warning: ‘f’ may be used uninitialized [-Wmaybe-uninitialized]
 1114 |                 n = f;
      |                 ~~^~~
dbodbc.c:1097:16: note: ‘f’ was declared here
 1097 |         double f;
      |                ^
dbodbc.c:1126:19: warning: ‘c’ may be used uninitialized [-Wmaybe-uninitialized]
 1126 |                 n = c;
      |                 ~~^~~
dbodbc.c:1095:14: note: ‘c’ was declared here
 1095 |         char c;
      |              ^
dbodbc.c:1133:16: warning: ‘n’ may be used uninitialized [-Wmaybe-uninitialized]
 1133 |         return n;
      |                ^
dbodbc.c:1096:14: note: ‘n’ was declared here
 1096 |         long n;
      |              ^
dbops.c: In function ‘lineFormatStack’:
dbops.c:191:40: warning: ‘dn’ may be used uninitialized [-Wmaybe-uninitialized]
  191 |                     (pdir == 'f' && dn == nullfloat) ||
dbops.c:101:16: note: ‘dn’ was declared here
  101 |         double dn;              /* double number */
      |                ^~
dbops.c:189:37: warning: ‘n’ may be used uninitialized [-Wmaybe-uninitialized]
  189 |                 if ((pdir == 'c' && !n) ||
      |                                     ^~
dbops.c:100:14: note: ‘n’ was declared here
  100 |         long n;
      |              ^
css.c: In function ‘cssPieces’:
css.c:1008:42: warning: ‘d2’ may be used uninitialized [-Wmaybe-uninitialized]
 1008 |                                 d2->next = d, d2 = d;
      |                                 ~~~~~~~~~^~~
css.c:802:31: note: ‘d2’ was declared here
  802 |         struct desc *d1 = 0, *d2, *d = 0;
      |                               ^~
In function ‘uncomment’,
    inlined from ‘cssPieces’ at css.c:815:2:
css.c:176:56: warning: ‘url0’ may be used uninitialized [-Wmaybe-uninitialized]
  176 |                                                 *w++ = *u;
      |                                                        ^~
css.c: In function ‘cssPieces’:
css.c:147:23: note: ‘url0’ was declared here
  147 |         char *w = s, *url0;
      |                       ^~~~
cc -I/usr/include/quickjs -Wdate-time -D_FORTIFY_SOURCE=2 -g -O2 -ffile-prefix-map=/tmp/edbrowse=. -fstack-protector-strong -Wformat -Werror=format-security -Wall -Wno-unused -D_FILE_OFFSET_BITS=64 -DEDBROWSE_ON_LINUX -I/usr/include/x86_64-linux-gnu  -g -ggdb -Wextra -c jseng-quick.c
In function ‘cursor_comm’,
    inlined from ‘syncup_table’ at dbops.c:2051:2:
dbops.c:1945:31: warning: ‘sortval2’ may be used uninitialized [-Wmaybe-uninitialized]
 1943 |                              ((sorttype == 'S'
      |                              ~~~~~~~~~~~~~~~~~
 1944 |                                && strcmp(sortstring1, sortstring2) < 0)
      |                                ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 1945 |                               || (sorttype != 'S' && sortval1 < sortval2)))) {
      |                               ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
dbops.c: In function ‘syncup_table’:
dbops.c:1835:23: note: ‘sortval2’ was declared here
 1835 |         int sortval1, sortval2;
      |                       ^~~~~~~~
In function ‘cursor_comm’,
    inlined from ‘syncup_table’ at dbops.c:2051:2:
dbops.c:1945:31: warning: ‘sortval1’ may be used uninitialized [-Wmaybe-uninitialized]
 1943 |                              ((sorttype == 'S'
      |                              ~~~~~~~~~~~~~~~~~
 1944 |                                && strcmp(sortstring1, sortstring2) < 0)
      |                                ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 1945 |                               || (sorttype != 'S' && sortval1 < sortval2)))) {
      |                               ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
dbops.c: In function ‘syncup_table’:
dbops.c:1835:13: note: ‘sortval1’ was declared here
 1835 |         int sortval1, sortval2;
      |             ^~~~~~~~
css.c: In function ‘qsaMatch’:
css.c:2387:28: warning: ‘rc’ may be used uninitialized [-Wmaybe-uninitialized]
 2387 |                         if (rc)
      |                            ^
css.c:2100:14: note: ‘rc’ was declared here
 2100 |         bool rc;
      |              ^~
In file included from /usr/include/quickjs/quickjs-libc.h:30,
                 from jseng-quick.c:18:
/usr/include/quickjs/quickjs.h: In function ‘JS_NewCFunctionMagic’:
/usr/include/quickjs/quickjs.h:969:34: warning: cast between incompatible function types from ‘JSValue (*)(JSContext *, JSValue,  int,  JSValue *, int)’ to ‘JSValue (*)(JSContext *, JSValue,  int,  JSValue *)’ [-Wcast-function-type]
  969 |     return JS_NewCFunction2(ctx, (JSCFunction *)func, name, length, cproto, magic);
      |                                  ^
cc -g -O2 -ffile-prefix-map=/tmp/edbrowse=. -fstack-protector-strong -Wformat -Werror=format-security -Wall -Wno-unused -D_FILE_OFFSET_BITS=64 -DEDBROWSE_ON_LINUX -I/usr/include/x86_64-linux-gnu  -g -ggdb -Wextra -Wdate-time -D_FORTIFY_SOURCE=2  -c -o ebrc.o ebrc.c
cc -g -O2 -ffile-prefix-map=/tmp/edbrowse=. -fstack-protector-strong -Wformat -Werror=format-security -Wall -Wno-unused -D_FILE_OFFSET_BITS=64 -DEDBROWSE_ON_LINUX -I/usr/include/x86_64-linux-gnu  -g -ggdb -Wextra -Wdate-time -D_FORTIFY_SOURCE=2  -c -o msg-strings.o msg-strings.c
jseng-quick.c: In function ‘setup_window_2’:
jseng-quick.c:3787:9: warning: ignoring return value of ‘asprintf’ declared with attribute ‘warn_unused_result’ [-Wunused-result]
 3787 |         asprintf(&wpc, "Wp`Set@%s", cf->fileName);
      |         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
jseng-quick.c: In function ‘set_basehref’:
jseng-quick.c:4448:17: warning: ignoring return value of ‘asprintf’ declared with attribute ‘warn_unused_result’ [-Wunused-result]
 4448 |                 asprintf(&wpc, "Wp`Set@%s", h);
      |                 ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
cc -g -O2 -ffile-prefix-map=/tmp/edbrowse=. -fstack-protector-strong -Wformat -Werror=format-security -Wall -Wno-unused -D_FILE_OFFSET_BITS=64 -DEDBROWSE_ON_LINUX -I/usr/include/x86_64-linux-gnu  -g -ggdb -Wextra -Wdate-time -D_FORTIFY_SOURCE=2  -c -o startwindow.o startwindow.c
jseng-quick.c: In function ‘embedNodeName’:
jseng-quick.c:2012:17: warning: ‘__builtin___strncpy_chk’ specified bound depends on the length of the source argument [-Wstringop-truncation]
 2012 |                 strncpy(b, nodeName, length);
      |                 ^
jseng-quick.c:2009:26: note: length computed here
 2009 |                 length = strlen(nodeName);
      |                          ^~~~~~~~~~~~~~~~
cc main.o buffers.o sendmail.o fetchmail.o html.o html-tags.o format.o stringfile.o ebrc.o msg-strings.o http.o isup.o css.o startwindow.o dbops.o dbodbc.o jseng-quick.o /usr/lib/x86_64-linux-gnu/quickjs/libquickjs.a -ldl -latomic -Wl,-z,relro -Wl,-z,now  -lcurl -lodbc -lpcre2-8 -lreadline -lssl -lcrypto   -lpthread -lm -lssl -lcrypto  -o edbrowse

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: uninitialized globals

[Tyler Spivey]

How do I reproduce this? I tried:

dpkg-buildflags --export=sh > export.out
source export.out
make clean && make

But edbrowse doesn't crash for me.

This is on a new Debian 12.5 machine.

On 4/5/2024 3:43 AM, Sebastian Humenda wrote:
> Hi
> 
> Pardon, I should have sent it to the list.
> 
> Karl Dahlke schrieb am 05.04.2024,  4:29 -0400:
>> the latest edbrowse 3.8.9 on Debian, compiled with GCC 12.2. It segfaults on
>> startup, and the backtrace suggests it is in main.c:1926, where it nzFree's
> […]
> […]
> 
> An excerpt of the build output is below this e-mail; it includes the build
> flags.
> 
>> I am rather stunned by this. I have been using C since 1980. Ever since the
>> original K&R, global uninitialized variables are 0. I don't understand how
> […]
> 
> I am bored by C standards and alike. I have been writing MISRA C for several
> years and I do remember the checkre requiring us to initialise explicitly
> every static (i.e. global) as its initialisation was either depending on the
> type or implementation defined, I do not recall.
> 
> Here is what nm says:
> 
> % nm src/edbrowse | grep 'sslCerts'
> 0000000000247088 D sslCerts
> 
> % nm src/edbrowse | fgrep allowJS
> % nm src/edbrowse | fgrep allowJS
> 
> 0000000000247079 D allowJS
> 
>> but that's not simple because there are dozens of global variables, and some
>> static variables too, that are not initialized, that I count on being zero.
>> If we got past sslCerts we would just run  into another one. We'd have to
> […]
> 
> Agreed. I actually just looked whether I can call up my.test.website and with
> this set, it worked :). I actually was trying out the QuicJS support.
> 
> Cheers
> Sebastian
> 
> === Build Output
> cc -g -O2 -ffile-prefix-map=/tmp/edbrowse=. -fstack-protector-strong -Wformat -Werror=format-security -Wall -Wno-unused -D_FILE_OFFSET_BITS=64 -DEDBROWSE_ON_LINUX -I/usr/include/x86_64-linux-gnu  -g -ggdb -Wextra -Wdate-time -D_FORTIFY_SOURCE=2  -c -o dbodbc.o dbodbc.c
> isup.c: In function ‘decodePostData’:
> isup.c:1276:30: warning: ‘c’ may be used uninitialized [-Wmaybe-uninitialized]
>   1276 |                         *w++ = c;
>        |                         ~~~~~^~~
> isup.c:1230:14: note: ‘c’ was declared here
>   1230 |         char c;
>        |              ^
> css.c: In function ‘hashSortCrunch’:
> css.c:3222:37: warning: ‘j’ may be used uninitialized [-Wmaybe-uninitialized]
>   3222 |                         mark->body[j++] = v->t;
>        |                                    ~^~
> css.c:3202:16: note: ‘j’ was declared here
>   3202 |         int i, j, distinct = 0;
>        |                ^
> css.c: In function ‘do_rules’:
> css.c:3106:9: warning: ‘tn’ may be used uninitialized [-Wmaybe-uninitialized]
>   3106 |         set_property_bool_t(tn, "inj$css", true);
>        |         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> css.c:2957:20: note: ‘tn’ was declared here
>   2957 |         const Tag *tn; // the text node that holds before or after text
>        |                    ^~
> dbops.c: In function ‘stringDate’:
> dbops.c:505:17: warning: ‘__builtin_strncpy’ output may be truncated copying 4 bytes from a string of length 7 [-Wstringop-truncation]
>    505 |                 strncpy(buf, buf + 4, 4);
>        |                 ^
> dbodbc.c: In function ‘retsFromOdbc’:
> dbodbc.c:937:30: warning: ‘input_length’ may be used uninitialized [-Wmaybe-uninitialized]
>    937 |                         rc = SQLGetData(hstmt, (ushort) (i + 1),
>        |                              ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>    938 |                                         c_type, q1, input_length,
>        |                                         ~~~~~~~~~~~~~~~~~~~~~~~~~
>    939 |                                         &output_length);
>        |                                         ~~~~~~~~~~~~~~~
> dbodbc.c:827:14: note: ‘input_length’ was declared here
>    827 |         long input_length, output_length;
>        |              ^~~~~~~~~~~~
> dbodbc.c:937:30: warning: ‘c_type’ may be used uninitialized [-Wmaybe-uninitialized]
>    937 |                         rc = SQLGetData(hstmt, (ushort) (i + 1),
>        |                              ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>    938 |                                         c_type, q1, input_length,
>        |                                         ~~~~~~~~~~~~~~~~~~~~~~~~~
>    939 |                                         &output_length);
>        |                                         ~~~~~~~~~~~~~~~
> dbodbc.c:826:15: note: ‘c_type’ was declared here
>    826 |         short c_type;           /* C data type */
>        |               ^~~~~~
> dbops.c: In function ‘stringTime’:
> dbops.c:636:28: warning: ‘c’ may be used uninitialized [-Wmaybe-uninitialized]
>    636 |                         if (c == 'A')
>        |                            ^
> dbops.c:593:14: note: ‘c’ was declared here
>    593 |         char c;
>        |              ^
> css.c: In function ‘cssParseLeft’:
> css.c:1336:52: warning: ‘sel2’ may be used uninitialized [-Wmaybe-uninitialized]
>   1336 |                                         sel2->next = sel, sel2 = sel;
>        |                                         ~~~~~~~~~~~^~~~~
> css.c:1258:21: note: ‘sel2’ was declared here
>   1258 |         struct sel *sel2;
>        |                     ^~~~
> dbodbc.c: In function ‘oneRetValue.constprop’:
> dbodbc.c:1114:19: warning: ‘f’ may be used uninitialized [-Wmaybe-uninitialized]
>   1114 |                 n = f;
>        |                 ~~^~~
> dbodbc.c:1097:16: note: ‘f’ was declared here
>   1097 |         double f;
>        |                ^
> dbodbc.c:1126:19: warning: ‘c’ may be used uninitialized [-Wmaybe-uninitialized]
>   1126 |                 n = c;
>        |                 ~~^~~
> dbodbc.c:1095:14: note: ‘c’ was declared here
>   1095 |         char c;
>        |              ^
> dbodbc.c:1133:16: warning: ‘n’ may be used uninitialized [-Wmaybe-uninitialized]
>   1133 |         return n;
>        |                ^
> dbodbc.c:1096:14: note: ‘n’ was declared here
>   1096 |         long n;
>        |              ^
> dbops.c: In function ‘lineFormatStack’:
> dbops.c:191:40: warning: ‘dn’ may be used uninitialized [-Wmaybe-uninitialized]
>    191 |                     (pdir == 'f' && dn == nullfloat) ||
> dbops.c:101:16: note: ‘dn’ was declared here
>    101 |         double dn;              /* double number */
>        |                ^~
> dbops.c:189:37: warning: ‘n’ may be used uninitialized [-Wmaybe-uninitialized]
>    189 |                 if ((pdir == 'c' && !n) ||
>        |                                     ^~
> dbops.c:100:14: note: ‘n’ was declared here
>    100 |         long n;
>        |              ^
> css.c: In function ‘cssPieces’:
> css.c:1008:42: warning: ‘d2’ may be used uninitialized [-Wmaybe-uninitialized]
>   1008 |                                 d2->next = d, d2 = d;
>        |                                 ~~~~~~~~~^~~
> css.c:802:31: note: ‘d2’ was declared here
>    802 |         struct desc *d1 = 0, *d2, *d = 0;
>        |                               ^~
> In function ‘uncomment’,
>      inlined from ‘cssPieces’ at css.c:815:2:
> css.c:176:56: warning: ‘url0’ may be used uninitialized [-Wmaybe-uninitialized]
>    176 |                                                 *w++ = *u;
>        |                                                        ^~
> css.c: In function ‘cssPieces’:
> css.c:147:23: note: ‘url0’ was declared here
>    147 |         char *w = s, *url0;
>        |                       ^~~~
> cc -I/usr/include/quickjs -Wdate-time -D_FORTIFY_SOURCE=2 -g -O2 -ffile-prefix-map=/tmp/edbrowse=. -fstack-protector-strong -Wformat -Werror=format-security -Wall -Wno-unused -D_FILE_OFFSET_BITS=64 -DEDBROWSE_ON_LINUX -I/usr/include/x86_64-linux-gnu  -g -ggdb -Wextra -c jseng-quick.c
> In function ‘cursor_comm’,
>      inlined from ‘syncup_table’ at dbops.c:2051:2:
> dbops.c:1945:31: warning: ‘sortval2’ may be used uninitialized [-Wmaybe-uninitialized]
>   1943 |                              ((sorttype == 'S'
>        |                              ~~~~~~~~~~~~~~~~~
>   1944 |                                && strcmp(sortstring1, sortstring2) < 0)
>        |                                ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>   1945 |                               || (sorttype != 'S' && sortval1 < sortval2)))) {
>        |                               ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> dbops.c: In function ‘syncup_table’:
> dbops.c:1835:23: note: ‘sortval2’ was declared here
>   1835 |         int sortval1, sortval2;
>        |                       ^~~~~~~~
> In function ‘cursor_comm’,
>      inlined from ‘syncup_table’ at dbops.c:2051:2:
> dbops.c:1945:31: warning: ‘sortval1’ may be used uninitialized [-Wmaybe-uninitialized]
>   1943 |                              ((sorttype == 'S'
>        |                              ~~~~~~~~~~~~~~~~~
>   1944 |                                && strcmp(sortstring1, sortstring2) < 0)
>        |                                ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>   1945 |                               || (sorttype != 'S' && sortval1 < sortval2)))) {
>        |                               ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> dbops.c: In function ‘syncup_table’:
> dbops.c:1835:13: note: ‘sortval1’ was declared here
>   1835 |         int sortval1, sortval2;
>        |             ^~~~~~~~
> css.c: In function ‘qsaMatch’:
> css.c:2387:28: warning: ‘rc’ may be used uninitialized [-Wmaybe-uninitialized]
>   2387 |                         if (rc)
>        |                            ^
> css.c:2100:14: note: ‘rc’ was declared here
>   2100 |         bool rc;
>        |              ^~
> In file included from /usr/include/quickjs/quickjs-libc.h:30,
>                   from jseng-quick.c:18:
> /usr/include/quickjs/quickjs.h: In function ‘JS_NewCFunctionMagic’:
> /usr/include/quickjs/quickjs.h:969:34: warning: cast between incompatible function types from ‘JSValue (*)(JSContext *, JSValue,  int,  JSValue *, int)’ to ‘JSValue (*)(JSContext *, JSValue,  int,  JSValue *)’ [-Wcast-function-type]
>    969 |     return JS_NewCFunction2(ctx, (JSCFunction *)func, name, length, cproto, magic);
>        |                                  ^
> cc -g -O2 -ffile-prefix-map=/tmp/edbrowse=. -fstack-protector-strong -Wformat -Werror=format-security -Wall -Wno-unused -D_FILE_OFFSET_BITS=64 -DEDBROWSE_ON_LINUX -I/usr/include/x86_64-linux-gnu  -g -ggdb -Wextra -Wdate-time -D_FORTIFY_SOURCE=2  -c -o ebrc.o ebrc.c
> cc -g -O2 -ffile-prefix-map=/tmp/edbrowse=. -fstack-protector-strong -Wformat -Werror=format-security -Wall -Wno-unused -D_FILE_OFFSET_BITS=64 -DEDBROWSE_ON_LINUX -I/usr/include/x86_64-linux-gnu  -g -ggdb -Wextra -Wdate-time -D_FORTIFY_SOURCE=2  -c -o msg-strings.o msg-strings.c
> jseng-quick.c: In function ‘setup_window_2’:
> jseng-quick.c:3787:9: warning: ignoring return value of ‘asprintf’ declared with attribute ‘warn_unused_result’ [-Wunused-result]
>   3787 |         asprintf(&wpc, "Wp`Set@%s", cf->fileName);
>        |         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> jseng-quick.c: In function ‘set_basehref’:
> jseng-quick.c:4448:17: warning: ignoring return value of ‘asprintf’ declared with attribute ‘warn_unused_result’ [-Wunused-result]
>   4448 |                 asprintf(&wpc, "Wp`Set@%s", h);
>        |                 ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> cc -g -O2 -ffile-prefix-map=/tmp/edbrowse=. -fstack-protector-strong -Wformat -Werror=format-security -Wall -Wno-unused -D_FILE_OFFSET_BITS=64 -DEDBROWSE_ON_LINUX -I/usr/include/x86_64-linux-gnu  -g -ggdb -Wextra -Wdate-time -D_FORTIFY_SOURCE=2  -c -o startwindow.o startwindow.c
> jseng-quick.c: In function ‘embedNodeName’:
> jseng-quick.c:2012:17: warning: ‘__builtin___strncpy_chk’ specified bound depends on the length of the source argument [-Wstringop-truncation]
>   2012 |                 strncpy(b, nodeName, length);
>        |                 ^
> jseng-quick.c:2009:26: note: length computed here
>   2009 |                 length = strlen(nodeName);
>        |                          ^~~~~~~~~~~~~~~~
> cc main.o buffers.o sendmail.o fetchmail.o html.o html-tags.o format.o stringfile.o ebrc.o msg-strings.o http.o isup.o css.o startwindow.o dbops.o dbodbc.o jseng-quick.o /usr/lib/x86_64-linux-gnu/quickjs/libquickjs.a -ldl -latomic -Wl,-z,relro -Wl,-z,now  -lcurl -lodbc -lpcre2-8 -lreadline -lssl -lcrypto   -lpthread -lm -lssl -lcrypto  -o edbrowse

Re: uninitialized globals

[Sebastian Humenda]

[-- Attachment #1: Type: text/plain, Size: 303 bytes --]

Hi

Tyler Spivey schrieb am 05.04.2024,  4:14 -0700:
>How do I reproduce this? I tried:
[…]

$ debcheckout edbrowse
$ cd edbrowse
$ git buildpackage
$ ./src/edbrowse https://scalable.capital

Do you need instructions that do not rely on devscripts and git-buildpackage?
Cheers
Sebastian

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: uninitialized globals

[Tyler Spivey]

Where do I get libquickjs? Do I need to be running Debian unstable?

On 4/5/2024 4:52 AM, Sebastian Humenda wrote:
> Hi
> 
> Tyler Spivey schrieb am 05.04.2024,  4:14 -0700:
>> How do I reproduce this? I tried:
> […]
> 
> $ debcheckout edbrowse
> $ cd edbrowse
> $ git buildpackage
> $ ./src/edbrowse https://scalable.capital
> 
> Do you need instructions that do not rely on devscripts and git-buildpackage?
> Cheers
> Sebastian

uninitialized globals

[Karl Dahlke]

And this captures the weirdness.

> % nm src/edbrowse | grep 'sslCerts'
> 0000000000247088 D sslCerts

It puts it in the .data section of the executable, I bet everyone else 
has B as I do, for .bss. 
Why does it not wind up in .bss? 
For grins I did the same with the object, before the linke, 
$ nm main.o | fgrep sslCerts 
00000004 C sslCerts 
It puts the uninitialized global in a C section, of some kind, maybe 
that's  a gcc thing, then at link time C becomes .bss, 
but in your world C becomes .data.

Karl Dahlke

Re: uninitialized globals

[Sebastian Humenda]

[-- Attachment #1: Type: text/plain, Size: 465 bytes --]

Hi

Karl Dahlke schrieb am 05.04.2024, 10:21 -0400:
>And this captures the weirdness.
>
>> % nm src/edbrowse | grep 'sslCerts'
>> 0000000000247088 D sslCerts
>
>It puts it in the .data section of the executable, I bet everyone else has B
>as I do, for .bss. Why does it not wind up in .bss? For grins I did the same
>with the object, before the linke, $ nm main.o | fgrep sslCerts 00000004 C

Also if you use the posted compiler and linker flags?

Cheers
Sebastian

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: uninitialized globals

[Tyler Spivey]

That worked, now I get a segfault.

On 4/5/2024 7:41 AM, Sebastian Humenda wrote:
> Hi
> 
> Tyler Spivey schrieb am 05.04.2024,  7:12 -0700:
>> I still can't get it to break. Here are all my steps:
> […]
> 
> Lovely, memory corruption keeps people so busy ;), thanks for investigating it.
> Would you please try this in your test env:
> 
> echo "certfile = /dev/null" > ~/.ebrc
> 
> Cheers
> Sebastian

uninitialized globals

[Karl Dahlke]

> Also if you use the posted compiler and linker flags?

Yes. I used same flags, as much as possible, save a couple of -I 
locations, 
sslCerts is in C the common symbols in main.o, as expected, but in B in 
my executable. 
No idea why it folds into D that is .data in your executable. 
Though I'm on a 32 bit machine so that could make a difference, and I'm 
sure a different version of gcc, 
but all that said, I've never heard of such a behavior, of putting 
those variables in the .data section.

Karl Dahlke

Re: uninitialized globals

[Tyler Spivey]

You patched sslCerts. That's causing your crash.

Without the patch, it should use whatever curl uses.

On 4/5/2024 7:54 AM, Tyler Spivey wrote:
> That worked, now I get a segfault.
> 
> On 4/5/2024 7:41 AM, Sebastian Humenda wrote:
>> Hi
>>
>> Tyler Spivey schrieb am 05.04.2024,  7:12 -0700:
>>> I still can't get it to break. Here are all my steps:
>> […]
>>
>> Lovely, memory corruption keeps people so busy ;), thanks for 
>> investigating it.
>> Would you please try this in your test env:
>>
>> echo "certfile = /dev/null" > ~/.ebrc
>>
>> Cheers
>> Sebastian
> 

Re: uninitialized globals

[Sebastian Humenda]

[-- Attachment #1: Type: text/plain, Size: 290 bytes --]

Hi

Tyler Spivey schrieb am 05.04.2024,  8:48 -0700:
>You patched sslCerts. That's causing your crash.

I didn't patch it, I specified a different source.

>Without the patch, it should use whatever curl uses.

It should not segfault, no matter what the configuration specifies.

Sebastian

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: uninitialized globals

[Dominique Martinet]

Sebastian Humenda wrote on Fri, Apr 05, 2024 at 08:18:15PM +0200:
> Tyler Spivey schrieb am 05.04.2024,  8:48 -0700:
> >You patched sslCerts. That's causing your crash.
> 
> I didn't patch it, I specified a different source.

I think that's what he means by patched:

edbrowse# cat debian/patches/ssl_certs 
Make edbrowse use the system CA certificates list.

---
 src/main.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

--- a/src/main.c
+++ b/src/main.c
@@ -34,7 +34,7 @@ bool errorExit;
 int webTimeout = 20, mailTimeout = 0;
 int displayLength = 500;
 int verifyCertificates = 1;
-char *sslCerts, *pubKey;
+char *sslCerts = "/etc/ssl/certs/ca-certificates.crt", *pubKey;
 int localAccount, maxAccount;
 struct MACCOUNT accounts[MAXACCOUNT];
 bool ebvar = true; // update environment variables before each shell command


That cannot be passed to nzFree as it's not an allocated value; this
patch should change to set it from strdup at start of main or something.
-- 
Dominique Martinet | Asmadeus

Re: uninitialized globals

[Sebastian Humenda]

[-- Attachment #1: Type: text/plain, Size: 641 bytes --]

Hi

Dominique Martinet schrieb am 06.04.2024,  3:37 +0900:
>Sebastian Humenda wrote on Fri, Apr 05, 2024 at 08:18:15PM +0200:
>> Tyler Spivey schrieb am 05.04.2024,  8:48 -0700:
>> >You patched sslCerts. That's causing your crash.
>> 
>> I didn't patch it, I specified a different source.
>
>I think that's what he means by patched:

Thanks :-). I also found that patch in the meantime. It's indeed the culprit.
I updated the patch, works now.
It also explains why sslCerts migrated to .data.

At least I now fixed two bugs in the packaging and got quickjs support Debian.
Let this be the announcement.

Cheers
Sebastian

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: uninitialized globals

[Dominique Martinet]

Sebastian Humenda wrote on Fri, Apr 05, 2024 at 08:47:14PM +0200:
> Thanks :-). I also found that patch in the meantime. It's indeed the culprit.
> I updated the patch, works now.
> It also explains why sslCerts migrated to .data.

Sorry I replied without thinking and suggested an equivalent patch, but
do you know why the curl default would not be good enough?

It looks like that patch originally came from this bug:
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=646802

but since then it shouldn't be needed anymore (at least I've never had
that problem on my builds without setting sslCerts - leaving it null
should use the system defaults), so the patch could probably be dropped.

There was discussion on irc to upstream the other two patches as well,
so we wouldn't need to patch the debian package; they both make sense to
me as well.

-- 
Dominique Martinet | Asmadeus